Experimental results obtained by AlexNet-based FCN.
\r\n\tHydrogen gas is the key energy source for hydrogen-based society. Ozone dissolved water is expected as the sterilization and cleaning agent that can comply with the new law enacted by the US Food and Drug Administration (FDA). The law “FDA Food Safety Modernization Act” requires sterilization and washing of foods to prevent food poisoning and has a strict provision that vegetables, meat, and fish must be washed with non-chlorine cleaning agents to make E. coli adhering to food down to “zero”. If ozone dissolved water could be successively applied in this field, electrochemistry would make a significant contribution to society.
\r\n\r\n\t
\r\n\tOxygen-enriched water is said to promote the growth of farmed fish. Hydrogen dissolved water is said to be able to efficiently remove minute dust on the silicon wafer when used in combination with ultrasonic irradiation.
\r\n\tAt present researches on direct water electrolysis have shown significant progress. For example, boron-doped diamonds and complex metal oxides are widely used as an electrode, and the interposing polymer electrolyte membrane (PEM) between electrodes has become one of the major processes of water electrolysis.
\r\n\t
\r\n\tThe purpose of this book is to show the latest water electrolysis technology and the future of society applying it.
Historically, progress in electron microscopy and its applications in virtually all branches of science and technology, including health care, was “instrument driven” in the initial decades until around the end of the previous millennium. During this time, efforts in developing instrumentation and methodology were concentrated primarily on improving the image resolution up to its physical limitations. If we restrict ourselves to scanning microscopy techniques, the goal was to minimize the dimensions of the spot of the primary beam incident on a sample. A second aim was to seek possibilities of combining the basic electron optical imaging tool with various analytical attachments. Tailoring the ultimate solution of the instrument to a specific microstructure- or nanostructure-bound task was not unfamiliar but certainly did not predominate. Progressively enhanced tools for the computer-aided simulation of the properties of electron optical elements and systems up to the complete simulation of the imaging process enabled the community of instrumentation scientists to turn to the design of a device according to the quantities and processes that are to be monitored and measured. A concentration on the deeper understanding of the collection of the image signal via identification of those parts of the energy and angular distributions of the emitted species that are acquired by detectors goes hand-in-hand with this. To put it more simply, understanding and interpretation of the image begins to overshadow the problems of generating and shaping the exciting primary electron beam. This kind of activity is surely facilitated if the energy of the electrons incident on the sample can be freely chosen in such a way that this degree of freedom can be fully utilized.
The range of energies of the primary electron beam available for the formation of sufficiently sharp beams in the scanning electron microscope (SEM) was for a long time limited by the practical obstacles that emerge when lowering the beam energy. These include the decreasing efficiency of extraction of electrons from the gun cathode, the increasing undesired manifestation of external spurious influences in proportion to the time of flight of the beam, and, most importantly, the enlargement of the size of the disks of confusion due to chromatic and diffraction aberrations making the primary spot dimension steeply extending toward low energies. However, since the very beginning of the electron microscopy era, the possibility of having an electron energy very low at the sample and sufficiently high in the column was known from devices using immersion objective lenses [1,2]. These traditionally appear in emission electron microscopes, where the sample itself is the source of electrons, mostly excited by incident photons, which are usually emitted at units or tens of electronvolts and strongly accelerated immediately above the sample surface biased to a high negative potential. Moreover, the idea of reverting the ray path and using the sample bias to retard the impinging primary beam generated and formed at a high energy in a standard electron gun and column is similarly old [3]. This idea appeared in the literature several times in the following years, though without serious attempts at implementation, with the exception of a study presenting the very first experimental results but nevertheless not at a convincing quality [4]. Surprisingly, the first successful implementation of the idea of retarding the electron beam just on the sample was not conditioned by preliminary simulation studies or the assembly of a dedicated device; enough courage to take on a task generally considered to have little prospects was sufficient [5,6]. The crucial objection met at the very beginning was the anticipated existence of various lateral fields close to the sample surface that would destroy the image by smearing both the primary and signal fluxes when they are so slow. Practical experience proved much less critical, and the first series of micrographs of a quality consistent through the full energy range down to 1 eV was published in 1993 [7].
The first decade of collecting experience with this innovative SEM mode was summarized in a review [8], and shorter reviews were then published at intervals of a few years [9,10]. More specific reviews concerning materials science and nanotechnology applications have appeared recently [11,12]. Here, we are to extend the series of reviews presenting in brief the instrumental and methodological fundamentals and listing the recent successful application examples.
When operating the scanning electron microscope (SEM) between, say, 5 and 10 keV, we enjoy an optimum compromise between conditions governing the primary beam formation and those characterizing the beam-sample interaction. We are able to extract sufficient electron current from all main types of sources and transport the beam over a column of tens of centimeters in length at a constant energy with a reasonable impact of spurious influences. Similarly, the corresponding wavelength of electrons and the energy spread of the beam in tenths of electronvolts, together with main aberration coefficients in millimeters, enable us to get a spot size of 1 nm or even slightly less. At the same time, the interaction volume of electrons of this speed in solids varies between units and hundreds of nanometers in size according to the sample material, although the true image resolution might be nearer to the spot size, provided sufficiently sharp structure details are present on the sample surface. Total yields of secondary and backscattered electrons are more or less comparable and can be easily separated as image signals. Conventional detectors acquire a part of the secondary electron (SE) emission attracted to a side-attached assembly with a front bias or a hollow cone of straight trajectories of backscattered electrons (BSE) with a ring-shaped diode or scintillator. The relief contrast of SE and material and channeling contrasts of BSE dominate routine SEM practice.
The observation of thin samples in a scanning transmission electron microscope (STEM) is traditionally performed at around 200 keV in dedicated instruments and around 30 keV in SEM devices equipped with a STEM attachment. While tissue sections for the former microscopes are prepared at hundreds of nanometers in thickness and samples containing heavier elements are correspondingly thinner, at tens of kiloelectronvolts, we have to go down to 100 nm and below. Detectors of transmitted electrons are split into several concentric rings so that the signal species can be sorted according to the scattering angle from the optical axis. This helps in obtaining sufficient contrasts from samples providing larger differences in local thickness or in atomic numbers of constituting elements. Of utmost importance here are the challenges of sample preparation, mainly as regards the aims of highlighting certain structure details considered interesting. Thanks to the concentrated illumination of primary electrons, samples for STEM can be somewhat thicker than those for TEM, but otherwise the experimentation issues are similar.
The traditional arrangement characterized above suffers from several drawbacks. Fast primary electrons interact mainly with atom cores, elastic forward scattering dominates, and inelastic scattering is weak. Scattering phenomena can be mostly explained within the laws of classical physics and the emitted signals do not respond to the quantum mechanical interior of the samples. The interaction volume of the beam in solid targets is too large, for which reason embedded structures such as precipitates are imaged fuzzy and both thin surface films and a great many relief details are invisible. The majority of signal species generated in depth do not penetrate to the surface to be emitted. Nonconductive materials charge up negatively to a surface potential in the kilovolt range. The edge effect, i.e. overbrightening of steeply inclined surface facets, impedes the measurement of distances and dimensions in SEM images. Fast BSE move along straight trajectories and the majority of them escape detection and impinge on the chamber walls. The main problems of STEM at standard energies include low contrasts, particularly with samples of living matter that need to be prepared with salts of heavy metals that highlight some but not all structure details, and the averaging of structure details across the sample thickness, which may hide some of them completely.
When decreasing the primary beam energy well below 5 keV or even to hundreds, tens, or units of electronvolts, we enter a different world of imaging conditions. Scattering events are increasingly dependent on the complete 3D potential distribution inside the target, i.e. they sense truncation of atomic potential by electrons and electron-electron interactions in general. Inelastic scattering and elastic backscattering become more important, so the crystallinity contribution to image contrast is enhanced. Below about 30 to 40 eV, electrons entering the target start behaving as Bloch waves, so their penetration inside is conditioned by the presence of unoccupied electron states in the particular direction and slow electron reflectivity can serve as a measure for the density of states and, therefore, can serve to fingerprint the orientation of grains in polycrystals, for example [13–15]. Interaction volume diminishes and information carried by emitted electrons is better localized laterally, while the surface sensitivity also increases, enabling one to observe all surface details and coverage. The edge effect disappears when the penetration depth stops exceeding the escape depth of signal species. Surface charging is reduced with increasing yield of SE, and at the “critical energy”, at which the fluxes of incident and outgoing electrons are identical, no charge is dissipated and noncharging microscopy is possible under any vacuum conditions. The total yield of signal electrons reaches its maximum somewhere in the range of hundreds of electronvolts, exceeding the unit level with minor exceptions. In this range, nonconductive samples charge positively, although charge balance is again achieved at the even lower first critical energy. The positive charging is moderate only, thanks to the partial retraction of the slowest emitted electrons by the field of the surface potential. The wavelength of electrons extends, and when it becomes similar to interatomic distances at units of electronvolts, interference of waves reflected on both sides of surface atomic steps (divided wavefront interference) or on both surfaces of ultrathin surface coatings (divided amplitude interference) becomes a new contrast mechanism. As in LEED, backscattering in the range of tens of electronvolts concentrates BSE to diffracted spots that may reveal the surface crystallinity when selectively detected. Moreover, electrons leaving the sample close above the surface barrier are partially reflected, so the height and shape of the barrier contribute to the imaging signal.
When reducing the primary beam energy in a STEM, we obtain, first and foremost, higher contrast. The differential cross-sections governing both the elastic and the inelastic scattering mechanisms are inversely proportional to the energy of electrons, mostly to energy squared. For this reason, the spatial density of the generation of the image information steeply increases so that, below 1 keV, we can get, for example, on tissue sections containing no agents with heavy metal salts not only very high contrast but also all structure details visualized, including those normally not highlighted with postfixation or staining media [16]. Naturally, the sections have to be thinner. For the range of hundreds of electronvolts, thicknesses below 10 nm are desirable. The production of such sections is, however, already feasible [17]. Moreover, thanks to reduced in-depth averaging of image layers, we get the chance of revealing tiny details unknown to date. With true 2D crystals of a single atom thickness, such as graphene, we can get nonnegligible electron penetration down to 1 eV [18] and obtain data that is extremely valuable for applications in nanoelectronics, for example.
The reasons for reducing the energy of electron incidence on targets in the scanning devices down to the lowest values, including the principle of implementation described below, have existed in the historical literature from the very beginning of the electron microscopy era and have been permanently demonstrated and supported more recently with indisputable results from the beginning of the 1990s. Starting from the 12th EUREM in Brno in 2000, the method was discussed at all large microscopy conferences and congresses in dedicated sessions. Still, the first commercially available SEMs allowing operation at tens of electronvolts did not appear until 2006/2007, and even then, the volume of application results in the literature remained extremely limited. Only quite recently, in the last 3 or 4 years, the “cathode lens” method has become a widely used standard tool in SEM practice. This circumstance starts erasing reasons for compiling review texts addressing just the method and transfers the observation material to reviews devoted to families of specimens. Nevertheless, the aim of this chapter is to summarize the fundamentals of the method, although it also features a list of successful applications in which very low energy SEM provides an important added value.
The problem of having very slow electrons on the sample surface and fast electrons in the column outside the sample was solved as early as 1932 with the immersion objective lens [1]. The first considerations counted on the sample as the source of electrons, but since 1942 a design has existed for an electron microscope delivering fast electrons toward the sample biased to a high negative potential in order to retard the electrons immediately before their impact on the sample surface [3]. The leading idea at that time was to reach energies of impact that produce the maximum yield of secondary electrons. For these reasons, there was really nothing new in the attempts to revitalize the principle that appeared sporadically over the following 50 years, though without convincing experimental results. With hindsight, we can now say that past attempts were probably not initiated with sufficient seriousness because of doubts connected with the necessity of heavy biasing of the specimen, which indicated conditions that were later shown to be not so strict, such as the necessity of having the sample surface very finely polished, similarly to the electrodes of electrostatic lenses.
The principle of implementation of very low energy SEM and STEM is simply as follows: the sample is biased to a high negative potential similar to the negative potential of the cathode in the electron gun of the microscope. The energy of electron impact on the sample is then given by the difference between the two potentials. An adverse factor here is the fact that, while the potential values are subtracted, their fluctuations are added, which amplifies the noise at the lowest impact energies quite substantially. It is much better to implement a “low voltage booster”, i.e. a positively biased axial tube insulated from the surrounding column that accelerates electrons just behind the gun anode and returns them to their original energy at the end of the objective lens [19,20]. The impact energy is then simply provided by another supply connected between the gun cathode and the sample. If we close the booster field at the end of the objective lens, we get a compound lens consisting of a (usually magnetic) focusing lens and an electrostatic retarding lens, an assembly long known [21] and incorporated into many commercially available electron microscopes under the name the Gemini lens.
The core concern is to know the electron optical parameters of the combination of the objective lens of the SEM with the electrostatic field above the sample. This field is, in fact, generated in a two-electrode immersion electrostatic lens with the sample serving as the cathode; this assembly is often called the “cathode lens” (CL). The anode of this lens has to have a central bore of some kind passing the primary beam so its field can be seen to be composed of a homogeneous retarding field and a field penetrating through the anode bore. On a three-electrode arrangement (cathode/grid/anode), Recknagel [2] derived an image formation theory employing expansion in ε/U, where ε is the electron energy in the surface plane of the sample and U is the acceleration voltage inside the lens. The analysis by Lenc and Müllerová [22,23], considering the anode field approximated with a second-order polynomial within a thin transition region in the anode plane, is more transparent. The main quantities, i.e. the spherical CS and chromatic CC aberration coefficients, were written (for large ratios EP/E) as [23]
where E is the landing energy of electrons, EP is the primary energy before retardation in the cathode lens, l is the length of the cathode lens field, D is the diameter of the anode bore, and CSf and CCf are the relevant aberration coefficients of the focusing objective lens. Obviously, the lower the landing energy of the electrons, the smaller the aberration coefficients, which is a relationship opposite to that valid for standard configurations without immersion, i.e. with energy independent aberrations. When decreasing the landing energy down to units of electronvolts, the aberration coefficients tend to fall in the micrometer range, which compensates very efficiently for other influences adverse from the point of view of image resolution. In Figure 1, we see typical energy dependences of the spot size for the sample bias switched on and off, in both cases with the beam aperture tailored separately to each landing energy, together with a particular case of the beam aperture before retardation fixed to a value optimum for a certain landing energy. At very low energies, the spot size deteriorates proportionally to E-3/4 in the standard SEM, while the cathode lens reduces this slope to E-1/4. For larger aberrations of the focusing lens, we can even get an energy-independent spot size [8].
We should mention here that, in the case of the Gemini compound lens, the working distance of the immersion electrostatic lens is nonzero (in contrast to the CL), so the basic aberration coefficients do not keep decreasing proportionally to the landing energy E without limitations, but the relevant equations contain an “absolute” member proportional to the working distance [24] so the spot size remains acceptable only down to a certain energy threshold. A comparison of the Gemini and cathode lenses is discussed in great detail in Ref. [25].
Energy dependences of the spot size for a typical magnetic focusing lens combined with a cathode lens.
The advantages of an assembly containing a cathode lens include the landing energy of electrons easily adjustable by the sample bias with the alignment of the microscope column untouched. Nevertheless, by altering the bias, we also vary the optical parameters of the electrostatic lens, first and foremost, the object distance of the sharp spot size and the dimensions of the field of view currently adjusted with scanning coils or electrodes. Approximate analytical calculations have shown these variations to be only moderate: at very low landing energies, the final magnification ranges between 1/2 and 2/3 of that with the CL off, in dependence on the position of the pivot point of the scanning system [7]. As regards the underfocusing necessary to compensate the cathode lens action, this amounts to units of micrometers per electronvolt at very low energies [7]. Algorithms suitable for correction of magnification and focusing in dependence on the cathode lens excitation can be found in Ref. [26]. Variations in the beam aperture and the beam impact angle with the landing energy are more important. In well-adjusted CL, we decrease only the axial component of the velocity of electrons, so the beam aperture enlarges, roughly in proportion to (EP/E)1/2. Similarly, amplification always occurs to the angle of impact of the primary beam on the sample surface when the rocking style of scanning is employed with a pivot point above the sample. Although the impact angle normally grows toward the margin of the field of view quite negligibly, after this “amplification”, it has to be considered, particularly when imaging crystalline samples exhibiting channeling contrast. At very low energies, we also have the field of view restricted by the total reflection of electrons taking place at a distance from the optical axis at which the glancing impact of electrons is achieved.
The cathode lens is aligned when the above-sample field is homogeneous and the nondeflected primary beam impacts normally on the sample. If any sample tilt is required, we have to count on some smearing of the primary spot due to the lateral field generated by the tilt. Moreover, the inclination of the beam by the mechanical tilt angle is again multiplied by the ratio of axial velocities before and after retardation. In this case, multiplication of the impact angle might prove advantageous. Practical experience has shown that all these apparent drawbacks of an assembly containing a cathode lens are manageable in practice without placing a great burden on the operator.
The alignment of the cathode lens consists of placing the anode bore on the optical axis and adjusting the sample by means of slight tilts to a parallel position with respect to the anode. Both these conditions are easily controllable with an anode made from a scintillator plate, preferably single crystal, sensitive on both sides. We then “see” the upper surface of the anode/detector and, at low magnification, can align its bore laterally to the screen center. At very low energies, we begin observing the above-sample assembly by means of electrons backscattered from the sample surface near the total reflection. Electrons emitted from individual sites on the sample are collimated in the strong field into narrow bundles so those emitted around the center of the field of view mostly escape through the anode bore. For this reason, we see the same anode bore dark from the bottom, and when both these circles are made concentric, the CL field is homogeneous. Reliable adjustment requires a sample stage equipped with two independent, mutually perpendicular sample tilts. A combination of sample tilt and rotation may also be helpful. The alignment is illustrated in Figure 2. The same approach may also be successful for other versions of the anode if the anode bore can be made visible around the view-field center.
Illustration of a well-aligned cathode lens: the bored scintillator disc of the BSE detector also serving as the anode (A), sample surface biased to a mirroring voltage (exceeding the acceleration voltage in the gun) (B), and mirror image of detector bore (the irregular contour is due to the unevenness of the sample consisting of a foil deposited on a mesh) (C).
The adverse external influences typical of the SEM include stray fields, primarily an external magnetic field. The deflection of the primary beam due to such a field depends on the time of flight of the beam, which decreases when reducing the range of flying at low energy. This is minimized in the case of zero working distance of the immersion lens, i.e. the cathode lens. When comparing the deflection at 20 keV and 200 eV, we get a ratio of 1:10 without the cathode lens but just 1:1.5 if the landing energy of 200 eV is obtained by retardation from 20 keV in the CL [25].
A strong electric field around the sample affects the primary electrons, although it affects the signal species even more because of their widely varying energy and directions of motion. Now, let us consider the negatively biased sample inserted between two grounded detectors, of which the upper one is either the CL anode or any detector above the anode to which electrons are passing through the anode bore and the lower detector is of an arrangement typical for the scanning transmission electron microscope (STEM) instrument. The homogeneous field of the CL accelerates the backscattered or transmitted electrons off the sample surface and collimates them into a bundle of a width decreasing with increasing field strength. The complete angular distribution for the full polar angle range (0, π/2) is concentrated in a circle of diameter [8].
where ke is the ratio of the final and initial energies of the emitted electrons. Generally, electrons are emitted at energies ranging between zero and the landing energy E, and although this interval may be quite short in comparison to the final energy, the factor ke may vary widely. We then get the signal electrons in the anode plane sorted according to their emission energy, which enables a certain energy filtering. In particular, secondary electrons (SE), if released at a low landing energy, are usually collimated to within a diameter of tenths of millimeters so they mostly escape through the anode bore, and if the anode also serves as the detector, we get an image signal composed almost exclusively of backscattered electrons (BSE). It is important that the complete BSE emission is usually collimated to a diameter in units of millimeters so we also acquire BSE emitted at large angles with respect to the surface normal that are usually omitted in conventional SEMs. As we will see later, SE also contribute to the transmitted electron (TE) signal acquired with a below-sample situated detector because they are accelerated similarly as TE passing the sample with various energy losses. Having the TE detector composed of traditional concentric sensitive detector rings for acquisition of bright-field (BF), dark-field (DF) and high-angle annular dark-field (HAADF) signal components, we should get SE in the BF channel only.
In modern SEMs, the sample is often immersed in a strong magnetic field of an open objective lens in order to improve the electron optical parameters of the column. When combining the open magnetic focusing lens with the CL, we face the question of the trajectories of signal electrons, particularly as regards the interpretation of their initial angular distribution on the basis of currents obtained in detector channels. As Figure 3 shows, the mismatch in the angular distribution normally appearing with the sample in a magnetic field is nearly fully eliminated by means of the electric field for ke = 11.
Trajectories of elastically backscattered and transmitted electrons from a sample immersed in the magnetic field of an open magnetic lens (a), from a sample to which the primary electrons are retarded 11 times (b), and from a sample surrounded by both fields (c).
As regards the detectors themselves, they are impacted by electrons accelerated approximately to the energy of electrons acquired in the microscope gun. The standard BSE detectors of SEMs, whether scintillator-based or semiconductor, are tailored to this energy, so no special precautions regarding detection are needed; every BSE detector installed in a particular SEM will work when we bias the sample and create a CL in its surroundings. Alternatives are listed in Ref. [8].
Slower electrons penetrate more shallowly into targets so the information depth of low-energy electron imaging shortens. Increased surface sensitivity provides enhanced information about the topography and possible coating of the surface, naturally including any contamination that may have concealed the surface to be examined. At hundreds of electronvolts and below, we enter the branch of instrumentation normally employed by surface physics methods such as low-energy electron diffraction, photoelectron, or Auger electron spectroscopy, in which ultrahigh vacuum (UHV) conditions are used and any sources of hydrocarbons have to be avoided. Experience has confirmed this assumption, although there is still the chance of improving surface cleanliness even under standard high (though dry) vacuum conditions where we have to remove primarily the adsorbed hydrocarbons loaded with the sample. Electrons at energies below 50 eV have proven themselves in removing the hydrocarbon molecules instead of decomposing them and creating the well-known carbonaceous contamination marking previous fields of view in the SEM [18,27,28]. This kind of cleaning caused, for example, the transmissivity of single-layer graphene to increase 2.5 times after bombardment with a dose of 1.3 Ccm-2 of 40 eV electrons [18].
Here we should provide answers to two frequently asked questions: How flat/smooth should the sample surface be? The CL field strength is usually about 2 or 3 kV/mm, i.e. 2 or 3 V/µm. Similarly strong lateral fields are generated with steeply inclined facets of surface unevenness details, with protrusions being more critical than dips. The tolerable relief height then depends on the landing energy we want to use. Protrusions in units of micrometers in height are well acceptable down to tens of electronvolts, whereas, at units of electronvolts, we can observe certain image deformations around surface steps around 1 µm in height. However, gently undulating surface structures with a p-p distance in units of micrometers are usually not apparent at all as disruptive damage to the image but are visible as surface topography details.
Can we image nonconductive samples? When placing a slab made from an insulating material on the conductive base of the sample holder and biasing the holder with respect to the surrounding ground, we get a situation equivalent to a capacitor with multilayer dielectric, composed here of a vacuum and a nonconductive sample slab. The potential drop then splits to both media according to their dielectric constants, so we get a somewhat lower negative bias on the sample surface than that led to the holder. However, when applying “good practice” in calibrating the landing energy scale, which consists of identifying its zero according to the disappearance of all topography details on a sample area as flat as available, we do not need to know the potential drop over the sample. What remains is to consider the landing energy scale modified according to the division between the sample and vacuum of every volt added to the holder bias, thereby giving the sample surface potential changed by less than 1 V. The potential drop across the sample can be detected when also checking the zero landing energy on the holder aside from the sample and the difference in both holder biases so determined can be used to correct the landing energy scale.
Since around a decade ago, the main producers of electron microscopes have been offering a cathode lens mode in their new instrument types, calling it beam deceleration, gentle beam, or decelerating optics. However, advances in the collection of new application results have been slow and seem to have been accelerating only quite recently as the Microscopy and Microanalysis Meeting 2015 has indicated. Owners of older SEMs are not completely excluded from enjoying the advantages of the CL mode in their instruments: simpler microscopes, more feasible adaptation; see Ref. [7]. The detection issue is solved by any functional BSE detector. All that is necessary is to insulate the sample from the stage and to connect it to a high-voltage feedthrough into the sample chamber. It is usually sufficient to insert the sample into a capsule made from a good insulating material and then load this capsule onto the stage. What remains is to connect the sample to the feedthrough in a way that does not cause havoc in the sample movements. More obstacles appear, of course, if the sample is loaded via an airlock. Nevertheless, plenty of positive experience was gathered with such adaptations in the 1990s.
Since the CL mode was first used for the practical examination of SEM samples, we have continued to map various families of samples in order to identify possible added value introduced by this mode. If the results of such a survey are to be efficiently demonstrated, it would seem to be natural to compare side-by-side micrographs taken at conventional energies at several kiloelectronvolts or more, with CL mode images differing not only in terms of the significantly reduced landing energy of electrons but usually also in terms of the collection of a broader or even complete angular distribution of backscattered electrons. In this chapter, we will follow this style of presentation for a selection of sample types often appearing in SEM practice.
The most straightforward expectation connected with decreasing the landing energy of primary electrons on the sample is their reduced penetration into the sample. Shortened information depth, along with reduced lateral diffusion, produces enhanced surface sensitivity, i.e. improved visibility of topographic details such as tiny dips, protrusions, and ridges, and also the sudden appearance of very thin surface coverage that is fully transparent and invisible at conventional energies. When comparing the two frames in Figure 4, we can identify examples of both these types of differences. Here, the penetration depth of primary electrons is the main factor; however, as shown below, the contribution of signal electrons from a broad range of polar angles of emission also plays a role.
Carbon nitride film 200 nm in thickness deposited on a silicon substrate covered by around 5-nm-thick native SiOx, delaminated due to compressive stress, CL mode, primary energy 9 keV.
The example shown in Figure 5 addresses a typical three-dimensional (3D) sample in which imaging of a highly complex surface structure requires the dimensions of the interaction volume of primary electrons in the material not exceeding the dimensions of the 3D details to be observed. The averaging of information over the volume from which BSE are capable of reaching the surface to be emitted smears out details smaller than, for example, hundreds of nanometers at 10 keV. Nevertheless, the interfaces between “bubbles” and vacuum are reproduced relatively sharply because the signal in pixels situated closely outside these bubbles does not include electrons diffusing through the material, so that the averaging is abruptly terminated at the margin.
Mesoporous carbon nitride foam as a carrier for catalytic gold nanoparticles, CL mode, primary energy 10 keV.
We might complete the paragraph concerning surface imaging with one problematic issue often mentioned where low electron energies are concerned. It has been said many times that decreasing the energy of incident electrons leads to reduced radiation damage of the sample. On one hand, every electron brings less energy, so certain inelastic events may no longer take place. On the other hand, the penetration depth shortens with decreasing energy faster than linearly, so dissipation of the delivered energy takes place in significantly shallower subsurface layer at growing spatial density of the dissipated power. From this point of view, the radiation damage connected with mechanisms active down to any particular low energy grows downward this energy. A well-known example is the creation of “black rectangles” marking the fields of view bombarded for a longer time with illuminating electrons. This carbonaceous contamination comes from the decomposition of adsorbed hydrocarbons under the impact of electrons. The more intense frames of these rectangles are caused by diffusion of the hydrocarbon molecules from the surrounding, so the phenomenon might be partly suppressed if we first immobilize the hydrocarbons around the next field of observation by electron impact. It is easy to verify that the intensity of this kind of contamination increases down to 100÷200 eV, i.e. to the fuzzy threshold of the shortest inelastic mean free path of electrons in solids, and only then fades out. Below some 50 eV, the radiation damage may usually be neglected.
In the previous paragraph, we presented samples with a heterogeneous or buckled or discontinuous surface coating and with a ragged spatial structure containing tiny protrusions and depressions. However, even when the sample is ideally flat and smooth and not covered with any thin layer, we may be confronted with imaging issues in cases in which very small objects are immersed just below the surface with the object tops lying on the same level as the neighboring surface. Good examples are precipitates in alloys prepared with an overall flat and smooth surface as shown in Figure 6. Here, we compare standard SEM micrographs at 10 keV in the BSE and SE signals. Precipitates are apparent in both frames, but most appear fuzzy. In the BSE image, the main contrast contribution comes from the atomic number difference, i.e. “material contrast”. The precipitates are composed of either Mg2Si or Mg3Si [30,31], so the difference of the mean atomic number with respect to that of Al is only 0.33 or 0.5, which is obviously sufficient for good contrast. The relatively sharp edges of the precipitates indicate an in-depth thickness much smaller than the information depth of the imaging, which reveals the margin information borne only by the BSE scattered in the narrow upper part of the pear-shaped interaction volume. In contrast to this, the SE appearance of the precipitates is fuzzier due to the contribution of the SE2 species released by BSE returning from the sample depth in a broad flux. The CL mode frame, thanks to much smaller interaction volume, not only shows quite sharp edges of the precipitates but also reveals their internal structure with a bright frame and dark core. The explanation of the internal structure obviously also has to incorporate the crystal structure of the precipitates, providing a specific contrast contribution when a sufficiently broad angular range of BSE is acquired, as we will show below.
Precipitates in Al-1.0 mass% Mg2Si with 0.4% excess Mg alloy, annealed, quenched, and age hardened: standard BSE image at 10 keV taken with a coaxial detector (a), SE image at 10 keV taken with a side-attached Everhart-Thornley detector (b), and the cathode lens image at 1,500 eV for 10 keV primary energy (c).
At high or medium electron energies, i.e. down to hundreds of electronvolts, the local crystallinity of the sample manifests itself in the channeling contrast. Incoming electrons are scattered at a rate proportional to the density of atomic planes they face, and if dense planes are inclined or even perpendicular to the surface, electrons penetrate into depth along interatomic channels and the probability of their backscattering decreases. For this reason, we get a BSE signal dependent on the local crystallinity. At hundreds of electronvolts, the scattering on atom cores begins to be combined with interaction with electrons in the target [32], and at even lower energies, below some 30 or 40 eV, the scattering is on a pure electron basis. If the incident electrons are of an energy close above the vacuum level, after gaining the inner potential, they appear within unoccupied energy bands that are already modified with the internal 3D potential distribution from the parabolic bands of free electrons and hence acquire a dispersion characteristic to the crystal system and its spatial orientation. Effectively, the impinging electrons convert into Bloch waves and move preferentially in directions of a high density of states. The reflectivity of electrons this slow then depends on the local density of states in the direction of impact of electrons on the crystal, i.e. electron states coupled to the incident electron wave. These have a surface-parallel wave vector component equal to that of the incident wave or differing in any surface reciprocal lattice vector [13,15,33].
We have already mentioned the enhanced crystal information contained in the BSE signal acquired at angles further from the surface normal. As we see in Figure 7, the polar-angle-sorted BSE imaging is composed of multiple contributions. The sample shown in this figure does not exhibit any local material contrast of details distinguishable at the given magnification, so we can compare contrasts of the grain orientation, grain boundaries, and the surface topography. Visibility of the grain boundaries dominates nearest to the optical axis; at higher angles and at angles near to 90°, we get the topography, whereas, between these angular intervals, the signal dependence on the grain orientation is most pronounced. Possible material contrast would appear at angles near to the optical axis. When assessing this figure, we should take into account that SE are also accelerated in the cathode lens field and then appear detectable very near to the optical axis. Their contribution is responsible for the edge effect visible in Figure 7a. Thus, as follows from Figure 7, the crystal contrast is best acquired at emission angles above 40° or 50° that are abandoned in conventional configurations of BSE detectors.
TRIP-aided bainitic ferrite (TBF) steel, imaged in the CL mode at 500 eV with primary energy of 4.5 keV, micrographs taken within the polar angle ranges of backscattered electrons: 0°–15° (a), 17°–26° (b), 28°–42° (c), 44°–61° (d), and 63°–90° (e).
Now, let us turn to the electron energy dependence of the crystal contrast. As Figure 8 shows, the acicular martensite structure is best visible around 500 eV, whereas, at kiloelectronvolt energies and also at tens of electronvolts, this kind of contrast is much weaker. The proposed explanation is as follows: at units of kiloelectronvolts, the image information is averaged within the depth range exceeding 100 nm, so the thin martensitic whiskers variously rotated with respect to the electron impact direction are averaged as regards their channeling ability, which reduces the resulting contrast. Moreover, this micrograph was taken with the cathode lens off, so only the BSE moving along straight trajectories within a rather narrow angular interval were acquired and the high-angle BSE providing more crystal contrast were abandoned. An optimum balance between the information depth and the thickness of the contrast producing structure details appears somewhere around 500 eV, and here, the full emission of the BSE is also detected. At 50 eV, the information depth falls below 1 nm, to which subsurface layer the structure features responsible for the contrast obviously do not raise. Having the information depth incorporating only two to three atomic layers, we may see only the surface reconstructed crystal that tends to convert differently oriented grains to rather unified structure. Another view might be based on a general fading of the channeling contrast at these energies.
Martensitic steel imaged in the CL mode with a primary energy of 6 keV.
Figure 9 refers to the above-mentioned contrast mechanism connected with the local density of electron states. We have to restrict ourselves to energies at which the absorption of hot electrons, proportional to the imaginary part of the crystal potential, is sufficiently low in order not to overwhelm the expected phenomenon [34]. As we see in the EBSD map, two of the Al grains in Figure 9, grains A and B, are near the orientation (111). Their reflectivity curves in Figure 9d are really quite similar, at least as regards the positions of the maxima and minima. The orientation of grain C is nearer to the middle of the color-coding stereographic triangle and the corresponding reflectivity dependence differs mainly below 20 eV. Grain D with an orientation near to (100) has a very different reflectivity. Obviously, having available reference curves for fundamental orientations of a particular crystal, we may be able to identify orientations of grains in a polycrystal of the same material [15]. As a potential alternative to the commonly used EBSD method, this approach would require an investment of decades of effort in data processing algorithms and tools, as was made during EBSD development. In principle, the low-energy reflectivity method offers better spatial resolution available at lower energy with a nontilted sample and faster data acquisition, thanks to only one-dimensional data collected for every pixel.
Identification of crystal grains in Al on the basis of reflectivity of very slow electrons: EBSD map (a), CL mode micrographs (b and c), and energy dependence of the reflectivity of selected grains (d) (reproduced from Ref. [29]).
In heavily deformed materials, some residual internal stress remains even after any deforming action is terminated. Consequently, elastic or plastic deformation of the crystal or of the grains in a polycrystal exists and should be observable in view of the importance of this parameter for material development and diagnostics. In relaxed grains, the BSE signal level revealing the penetration of primary electrons along the interatomic channels is constant over the grain area up to the grain boundary where dislocations and other defects are normally clustered. If microscopic deformations exist inside a grain, the conditions for the channeling of electrons vary even inside the grain, which leads to signal variations proportional to the extent of deformation and correspondingly distributed. Gradual signal variations without abrupt changeovers are characteristic, as Figure 10 demonstrates on a margin of the Vickers indent. On the inclined wall of the indent, we may observe plastically deformed grains of internally varying brightness owing to changes in crystallographic orientation, whereas, outside the indent, the grains are of a constant signal over their full area.
Margin of a Vickers indent in an annealed polycrystalline Cu sample, CL mode at 500 eV showing the residual stress distribution inside the indent (left) and relaxed grains (right).
Figure 11 demonstrates the dependence of the visibility of grain deformation on the energy of electrons. Similarly as in Figure 8, we have the contrast culminating around 500 eV and nearly invisible at the nonreduced primary beam energy or at 50 eV. The reasons for this phenomenon are very probably identical to those regarding contrasts between relaxed grains, as discussed in the previous paragraph.
X210Cr12 ledeburitic steel heated to a semisolid state, heavily deformed and cooled, CL mode micrographs for the primary energy of 6 keV.
The measurement of the local density of active dopants in semiconductors is of crucial importance for semiconductor technology in its all versions and phases, particularly as regards integrated circuits. While silicon wafers recently began achieving 45 cm in diameter, the characteristic dimensions of circuit structure details have been stepwise diminished from 45 to 32 nm and then to 22 nm, allegedly with 16 nm as the future prospect. The gigantic number of doped patterns excludes any true interoperational checks during production, but the possibility of measuring both the critical dimensions and true local density of a dopant still poses a difficult task for instrumentation designers. Thanks to its nondestructive application, the flexible size of the field of observation, resolution below 1 nm, and possible response to all sample characteristics including topography, chemistry, crystallinity, and electronics, SEM is traditionally considered the most suitable diagnostic method. However, important obstacles have hindered the development of this SEM application. The measurement of the critical dimensions of patterns faces the edge effect, i.e. overbrightening of the margins of surface steps imaged with secondary electrons that are also emitted from the sidewall as an extended surface. The edge effect disappears when the penetration depth of primary electrons drops beneath the escape depth of SE; with silicon sample, this happens at a landing energy of around 300 eV [35], which requires the use of the CL mode. Although the visibility of doped patterns was verified long ago [36], the explanation of dopant contrast has fluctuated between a great many alternatives. A review of this topic up to 2000 [37] considered the governing factor to be the difference in ionization energy between p- and n-types and its balancing by above-surface electric “patch fields”. Later works incorporated below-surface fields on the junction between semiconductor and carbonaceous contamination [38] and the filtering action of the surface potential barrier [39]. The first observations of doped semiconductors in the CL mode [40] provided data about the contrast dependence on the external electric field of the CL and on vacuum conditions and the configuration of the detector. An explanation of these observations relied on the influence of surface passivation and the CL field causing the dopant contrast enhanced with respect to conventional SEM imaging.
This chapter does not aim to make a decision about the proposed models of the dopant contrast mechanism but merely to gather the most important experimental data. In order to avoid the low repeatability of data recorded in the past on series of samples with only one density of dopant per sample, which was to a certain extent the result of nonidentical surface treatment causing differences in the potential barrier, our samples were prepared with patterns doped by four different dopant densities. Figure 12 confirms the main relation known for a long time, namely, the higher imaging signal from p-type patterns compared to n-type patterns. As we see, the dopant contrast increases from 4 keV to 1 keV. At 4 keV (and primary energy 7 keV), SE are accelerated to 3 keV before detection, which is not sufficient to obtain a high signal from a scintillator covered with a metallic layer. This change of landing energy also causes acquisition of a broader flux of BSE, which are collected completely at 1 keV. Simultaneously, the slowest SE escape detection through the central detector bore, although faster SE still contribute to the image. However, this finding indicating an increase in contrast due to the loss of the slowest SE somewhat contradicts the measurement of the p/n contrast made with energy filtering of the SE in which the contrast carried by SE up to 10 eV was higher than that for the full SE emission [41]. In our case, SE below 3 eV escape detection fully, whereas, at higher energies, the central parts of the collimated SE flux are not detected. An important factor is that, at around 1 keV, the image contrast is reliably proportional to the dopant density and can be quantified for the sake of density measurement.
When going down with landing energy of electrons, the dopant contrast in Figure 12 diminishes and at tens of electronvolts inverts, giving the p-type patterns darker than the n-type background. No explanation has yet been proposed for this phenomenon, connected solely with the BSE emission because SE completely escape through the detector bore. Down to units of electronvolts, this inversed contrast is enhanced, but its dependence on the dopant density is much weaker. Near to zero energy, one more contrast inversion appears and we again get brighter p-type silicon. This effect is discussed below.
p-type doped patterns of dopant concentrations: 1019 cm-3 (B), 1018 cm-3 (C), 1017 cm-3 (D), and 1016 cm-3 (E) on an n-type substrate (1015 cm-3, A), etched in the Buffered Oxide Etch, CL mode, primary beam energy 7 keV, current 500 pA.
It is no great surprise that the “opposite” structure of n-type patterns on a p-type substrate in Figure 13 exhibits an opposite brightness relation with darker strips. Again, we have the contrast increasing toward 1 keV, now with a less pronounced proportionality between the contrast and the dopant density. At tens of electronvolts, the contrast inverts again and becomes more dependent on the dopant density. No important change takes place near the mirror image.
n-type doped patterns of dopant concentrations: 1019 cm-3 (B), 1018 cm-3 (C), 1017 cm-3 (D), and 1016 cm-3 (E) on a p-type substrate (1015 cm-3, A), etched in the Buffered Oxide Etch, CL mode, primary beam energy 6 keV, current 500 pA.
Important findings concern the influence of the surface status on the observed contrasts. Particularly, under standard vacuum conditions, when carbonaceous contamination is always present, we observe contrast dependence on the rate of contamination, impact angle of primary electrons, electron dose, detector geometry and position, etc. [42]. However, even the mere storage of an originally etched sample in air for several weeks leads to inversion of the contrast in the UHV microscope (see Figure 14).
n-type patterns on a p-type substrate (see Figure 13) after etching in the Buffered Oxide Etch (a) and after several weeks in air (b).
Now, let us return to the ultimate inversion of the p/n contrast on a p-type pattern/n-type substrate sample near zero landing energy. Here, we obtain an extremely high signal from p-type patterns though with variously shaped black dots (Figure 15). The measurement shows that we are getting a “full” contrast here between the total beam current and no current. The explanation was based on the injection of electrons in the p-type patterns and their recombination and on the ability of the small surface charge around 1 V thereby created to decrease the very low landing energy of incident electrons near enough conditions of total reflection [43]. The total reflection of the primary electron beam is directed toward the above-sample detector based on the bored scintillator disc, so we get white pixels where the reflected ray hits the scintillator surface and a black area where it hits the bore.
Total reflection phenomena on p-type doped patterns: lateral shifts of the pattern group influencing which pixels illuminate the active detector area around the central bore (a–d); dependence of pattern charging on the landing beam current: 30 pA (e); 100 pA (f); 600 pA (g); and 1.5 nA (h); and dependence on the electron dose proportional to the pixel dwell time: 560 ns (i); 1.76 µs (j); 5.36 µs (k); and 48.56 µs (l).
In Figure 15, we see the phenomenon dependent on the mutual position of the group of patterns and the detector and also on the beam current and dose. Obviously, the emergence of this phenomenon depends on a certain relationship between the electron dose sufficient to cause it and dopant density, which may offer another tool for measuring this density.
An extremely important family of microscopic samples consists of thin sections of tissues or more generally of live matter and various organic materials. These samples are usually amorphous and composed of light elements. For microscopic observation, they are cut into layers of submicrometer thicknesses and observed in TEM or STEM, traditionally using electrons in a range of hundreds of kiloelectronvolts and more recently of tens of kiloelectronvolts. An exception to this is low-voltage TEM operated at around 5 keV [44,45]; the success of this instrument is based on partially overcoming difficulties with the achievement of a sufficient image contrast in structures composed of light elements. In order to get good image contrast, sophisticated preparation procedures have been elaborated, particularly for tissue sections. These include fixation in various agents, immersion in resins, postfixation with osmium tetroxide, and staining with agents such as uranyl acetate or lead citrate. High contrast is then obtained where the heavy metal species from the chemicals used in preparation are located. Unfortunately, only some details of the structure are highlighted in this way, mainly with staining. When using STEM equipped with a CL mode and decreasing the landing energy of electrons, we observe a dramatic increase in contrast even when no heavy metal-containing substances are used in preparation [16] (see Figure 16).
Section of mouse heart muscle, not fixed with osmium tetroxide and not stained: 10 nm section imaged by conventional TEM at 80 keV (a), CL mode micrograph taken at 500 eV with a primary energy of 4.5 keV (b), and energy dependence of the contrast of this sample in the CL mode (c).
An important finding made in a pioneering study [16] concerns changes in the embedding resin. It has been known that the resin is partially “radiation damaged” under electron bombardment with the consequence of the increased electron transmissivity of sections, but this increase was moderate only. The rate of this effect was compared at various electron energies, and whereas 5 keV STEM increases the transmitted flux two times after a dose of about 5×10-3 Ccm-2, at 500 eV, the increase was 20× at a doubled dose [16]. The idea is that, under electron impact, the resin is partly depolymerized and monomers are then released from the surface; no losses in the observable structure details have been discovered after this “ultimate” preparation step performed in situ.
Sections for ultralow-energy STEM have to be very thin, preferably below 10 nm [17]. These are transparent down to 500 eV if “electron bombardment thinning” is utilized and micrographs have even been taken at tens of electronvolts. The advantages here include averaging of image details over a much shorter trajectory across the layer, which may reveal some features, usually bright spots, not observable in thicker sections at any electron energy.
The application of ultralow-energy STEM in polymers, polymer blends, and composites forms a branch to be examined next.
Intuitively, crystalline layers should be more transparent for slow electrons than amorphous sections because of the expected possibility of channeling among atomic planes or columns. For this reason, ultralow-energy STEM with the cathode lens mode has been applied to 2D crystals, firstly to graphene. The first results were achieved when comparing the Raman spectroscopy identification of flakes of a certain thickness with STEM observation [46]. Even flakes exhibiting a Raman spectrum corresponding to single-layer graphene were found to be composed of tiny flakes of various thicknesses and small holes. This finding argues in favor of introducing CL-mode STEM as an acknowledged method for the diagnostics of graphene and other 2D crystals.
Because the available graphene samples are generally composed of flakes only rarely exceeding micrometer sizes and usually overlapping each other at least partly, we need first and foremost to be able to obtain sufficient contrast between sites differing in thickness by a single layer of carbon atoms. As we see in Figure 17, this demand is met at about 100 eV in the transmission mode, whereas this kind of contrast is not available in the BSE signal.
CVD graphene samples deposited on lacey carbon lying on a copper mesh, commercially available sample declared as three- to five-layer graphene: reflected signal for 1 keV (a) and 100 eV (b) landing energies and transmitted signal for 1 keV (c) and 100 eV (d).
One sort of commercially available graphene is a polycrystalline, CVD prepared material deposited on lacey carbon with eyes of up to 1 or 2 µm in diameter, which was used in our study. In Figure 18, we see the layer-by-layer contrast at higher magnification in a standard vacuum microscope, together with a detailed frame showing the polycrystalline structure of individual domains, though of continuous single-layer graphene (1LG hereinafter). These samples have been used for the measurement of the electron transmissivity through various numbers of graphene layers, assuming that the total area of nucleation centers visible in Figure 18a is negligible when averaging the transmitted signal over the lacey eye, and similarly, the existence of the domains does not influence the result.
CVD graphene (see Figure 17) in transmitted electrons at higher magnification, 220 eV (a) and 1 keV (b), and at lowest energies, 40 eV (c) and 4 eV (d).
We can notice in Figure 18d the electron penetration through the sample even at 4 eV. In fact, transmitted current was measurable down to or even below 1 eV. This enabled us to measure the transmissivity across the full energy scale from kiloelectronvolts to 1 eV. The result in Figure 19 was surprising in view of the expected increase in the inelastic mean free path of electrons below about 50 eV, which is not only generally observed [47] but also results from calculations [48]. The explanation should be sought in the fact that graphene in flakes sufficiently exceeding units of nanometers in size does not have an energy gap between the valence and conductance bands meeting at the Dirac point [49]. This causes interband transitions as the inelastic scattering mechanism working down to arbitrarily low energy losses in scattering events.
Energy dependence of the transmissivity of slow electrons through single-layer graphene.
In Figure 19, we see the transmissivity apparently exceeding 100% above 200 eV. This effect is caused by SE released inside the sample with original direction of movement toward the bottom surface where they are emitted and accelerated toward the detector together with the transmitted primary electrons. The excess current is then balanced from the earth. A detailed measurement of the electron transmissivity has been performed for up to 7LG. The ratio of transmissivities 1LG/7LG was found to be largest at 40 eV (over 6), so this energy is recommended for the reliable counting of graphene layers [18].
Electron microscopy of graphene below 10 eV was also examined in the reflection mode in a low-energy electron microscope (LEEM) [50]. The BSE signal was found oscillating in such a way that n layers of graphene (grown on various surfaces or free standing) produce n-1 minima in the reflectivity between 0 and 8 eV [50,51]. We have also verified this relation in STEM and were even able to confirm experimentally the second band of oscillations predicted by calculations in the range between 13 and 20 eV [27]. Oscillations are demonstrated in Figure 20 on a stack of multiple layers of graphene.
Multilayer graphene deposited by the CVD technique on a Cu foil, CL mode micrographs taken for the primary energy 6 keV.
Finally, let us return to the “radiation damage” created by very slow electrons, mentioned above in connection with the decomposition of the resin in which the tissue section was embedded. Electrons below 40 or 50 eV were found to effectively clean the graphene samples increasing their transmissivity: a dose of 1.5 Ccm-2 of 40 eV electrons increased the transmissivity of 1LG 2.5 times [18]. With graphene as a single atomic layer, there are no possible doubts about the mechanism of this radiation damage — if the sample is not bored, the only alternative is the removal of adsorbed gases, which was observed even under standard vacuum conditions. We suppose the removal of the hydrocarbon molecules instead of their decomposition normally generating the carbonaceous contamination, as happens immediately when increasing energy to, say, 200 eV. This kind of in situ surface cleaning promises a method for some surface studies performed outside UHV.
In Figure 21, the cleaning of graphene with slow electrons is shown on decreased reflectivity and increased transmissivity of longer bombarded parts of the fields of view.
Demonstration of the cleaning effect of bombardment with very slow electrons, single-layer graphene, CL mode at 30 eV: original state shown in reflected (a) and transmitted (d) signal, decreased reflectivity (b), and increased transmissivity (c) after impact of about 1 Ccm-2 of 30 eV electrons.
An observation of graphene samples grown on substrates at very low energies also makes it possible to distinguish between overlayer and underlayer growth of the second layer. A decision is made according to the comparison of micrographs taken at few hundreds of electronvolts, when all the “wedding cake” of stacked layers is visible, with a frame at units of electronvolts when we see only the topmost layer [27].
During a little more than 20 years of laboratory existence and nearly a full decade since the introduction of a commercial device, the cathode lens mode has slowly ceased to be considered a methodological novelty. If any future review of the method is to be written, its content will probably be limited to a particular type of samples and place the emphasis on models of contrast mechanisms rather than the mere comparison of the traditional and new appearance of micrographs. However, this time the older approach has been chosen. We have seen above that not all effects observed even when examining quite common samples have already been satisfactorily explained. This chapter aims to encourage colleagues capable of creating the desired physical models.
This chapter mostly summarizes the results of research performed by the Group of Microscopy and Spectroscopy of Surfaces at the Institute of Scientific Instruments of the CAS in Brno, headed by Dr. Ilona Müllerová, who began CL mode implementation in an SEM and continuously manages the project. The presented data and micrographs have mostly been collected by former Ph.D. students of the group: Dr. Miloš Hovorka, Dr. Šárka Mikmeková, Dr. Eliška Mikmeková, Dr. Filip Mika, Dr. Zuzana Pokorná, and Dr. Ivo Konvalina. Thanks are also due to Dr. Aleš Paták, Mr. Jiří Sýkora, and Mr. Pavel Klein. Some of the samples have been provided by Dr. Jana Nebesářová (Biology Centre, České Budějovice), Professor Kenji Matsuda (University of Toyama, Japan), Professor Bohuslav Mašek (University of Western Bohemia, Plzeň), and Dr. Baowen Liu (UNIST, Ulsan, South Korea). In recent years, the program has been supported by the project TE01020118 of the Technology Agency of the Czech Republic (Competence Centre: Electron Microscopy) and, in part, by MEYS CR (LO1212).
Robotics, as a commercial technology, started to be widespread some decades ago, but instead of decreasing, it has been growing year by year with new contributions in all the related fields that it integrates. The introduction of new materials, sensors, actuators, software, communications and use scenarios converted Robotics in a pushing area that embraces our everyday life. New robotic morphologies are the most shocking aspect that society perceives (i.e., the first models of each type generally produce the largest impact), but the long-term success of robotics is found in its capability to automate productive processes. Manufacturers and developers know that the market is found not only in large-scale companies (car manufacturers and electronics mainly) but also in the SME that provides solutions to problems that are manually performed so far. Also, robotics has opened the doors to new applications that did not exist some years ago and are also attractive to investors. These facts, together with lower prices for equipment, better programming and communication tools, and new fast-growing user-friendly collaborative robotic frameworks, have pushed robotics technology at the edge in many areas.
It is clear that industrial robotics leads the market worldwide, but social/gaming uses of robots have increased sales. Nevertheless, the most promising scenario for the present time and short term is the use of robots in commercial applications out of the plant floor. Emergency systems, inspection, and maintenance of facilities of any kind, rescues, surveillance, agriculture, fishing, border patrolling, and many other applications (without military use) attract users/clients because their use increases the productivity of the different sectors, low prices and high profitability are the keys.
There exist many robot morphologies and types (surface, underwater, aerial, underground, legged, wheels, caterpillar, etc.) but authors want to draw attention in the unmanned aerial vehicles (UAVs), which have several properties that make them attractive for a set of application that cannot be done with any other type of robot. First, those autonomous robots can fly, and therefore, they can reach areas that humans or other robots cannot. They are light, easy to move from one area to another, and can be adapted to any area, terrain, soil, building, or facility. The drawback is the fragility in front of adverse meteorological events, and their autonomy is quite limited compared with unmanned surface vehicles (USVs).
UAVs have seen the birth of a new era of unthinkable cheap, easy applications up to now. The authors would like to focus its use in the maintenance and inspection of industrial facilities, but specifically in the inspection of pipes in big, complex factories (mainly gas and oil companies) where the manual inspection (and even location and mapping) of pipes becomes an impossible task. Manned helicopters (with thermal engines) cannot fly close to pipes or even among a bunch of pipes. Scaffolds cannot be put up in complex, unstable, and fragile pipes to manually inspect them. Therefore, a complex problem can be solved through the use of UAVs for inspecting pipes of different diameters, colors, textures, and conditions in hazardous factories. This problem is not new and some solutions have been brought to an incipient market. Works as those in [1, 2] propose the creation of a map of the pipe set navigating among it with odometry and inertial units [3]. Obstacle avoidance in a crowded 3D world of pipes becomes of great interest when planning a flight; in [4], some contributions are made in this direction although the accuracy of object is deficient to be a reliable technology. Work in [5] overcomes some of the latter problems with the use of a big range of sensors, cameras, laser, barometer, ultrasound, and a computationally inefficient software scheme made the UAV too heavy and unreliable due to the excessive sensor fusion approach.
Many of the technical developments that have helped robotics grow have had a wider impact, especially those related with increasing computational power and parallelization levels. Faster processors, with tens of cores and additional multiple threat capabilities, and modern GPUs (graphics processing unit) have led to the emergence of GPGPU (general-purpose computing on GPU). These type of computing techniques have led to huge advances in the artificial intelligence (AI) field, producing the emergence of the “deep learning” field. The deep learning (DL) field is focused in using artificial neural networks (ANNs) that present tens or hundreds of layers, exploiting the huge parallelization capabilities of modern GPU. This is used in exploiting computational cores (e.g., CUDA cores), which compared on a one-to-one basis with a processor core, they are less powerful and slower, but can be found in amounts of hundreds or thousands. This has allowed the transition from shallow ANN to the deeper architectures and innovations such as several types of convolutional layers. In this work, the authors present a novel approach to detect pipes in industrial environments based in fully convolutional networks (FCNs). These will be used to extract the apparent contour of the pipes, replacing most of the architecture developed in [6] and discussed in Section 2. To properly train these networks, a custom dataset relevant to the domain is required, so the authors captured a dataset and developed an automatic label generation procedure base in previous works. Two different state-of-the-art semantic segmentation approaches were trained and evaluated with the standard metrics to prove the validity of the whole approach. Thus, in the following section, some generalities about the pipe detection and positioning problem are discussed, and the authors’ previous work [6] on it, as it will be relevant later. The next section discusses the semantic segmentation problem as a way to extract the apparent contour, both surveying classical methods, considered for earlier works, and state of the art deep-learning-based methodologies. The fourth section describes how the automatic label generator using multimodal data was derived and some features to the process. The experimental section starts discussing the metrics employed to validate the results, the particularities of the domain dataset generated and describes how an AlexNet FCN architecture was trained through transfer learning and the results achieved. To conclude, some discussion on the quality of the results and possible enhancements is introduced, discussing which would be the best strategies to follow continuing this research.
As it has been discussed, inspection and surveying are a frequent problem where UAV technologies are applied. The most common scenario found is that of a hard to reach infrastructure that is visually inspected through different sensors onboard a piloted UAV. Some projects have proposed the introduction of higher level perception and automation capacities, depending on the specific problem. In these cases, it is common to join state-of-the-art academic and industrial expertise to reach functional solutions.
In one of these projects, the specific challenge of accurately detecting and positioning a pipe in real time using only the hardware deployable in a small (per industry standards) UAV platform was considered (Figure 1), with several solutions studied and tested (including vision- and LIDAR-based techniques).
One of the UAV used for the development of perception tasks in the AEROARMS project. Several sensors were deployed, processing them with a set of SBCs (single-board computers), including a Velodyne LiDAR, two different cameras, ultrasonic range-finder (height), and optical flow.
In the case of LIDAR-based detection, finding a pipe is generally treated as a segmentation problem in the sensor space (using R3 data collected as “point clouds”). There are many methods used for LIDAR detection, but the most successful are based on stochastic model fitting and registration, commonly in RANSAC (Random Sample Consensus [7]) or derived approaches [8, 9]. Three different data density levels were tested using the libraries available through ROS: using RANSAC over a map estimated by a SLAM technique, namely LOAM [10]; detecting the pipe in a small window of consecutive point clouds joined by an ICP-like approach [11]; and finally to simply work using the most recent point cloud. The first approach probed to be computationally unfeasible, no matter what optimization was tested, as even working with a single datum cloud point could be prohibitive if not done carefully. To enhance the performance, the single cloud point approach was optimized employing spatial and stochastic filtering to reduce the data magnitude, and a curvature filter allowed to reduce fake positives in degenerate configurations, producing robust results at between 1 and 4 Hz. To solve the same problem with visual sensors, a two-step strategy was used. In order to estimate the pose of the pipes to be found, they were assumed to be circular and regular enough to be modeled as a straight homogeneous circular cylinder. This allowed using a closed-form conic equation [12], which related the axis of the pipe (its position and orientation as denoted in Plücker coordinates) with the edges of its projection in the image space. While this solves the positioning problem, the detection probed to be a little more challenging: techniques based on edge detection, segmentation, or other classical computer vision methods used to work under controlled light but failed to perform acceptably in outdoor scenarios. This issue was solved by introducing human supervision, where an initial seed for the pipe in the image sensor space was provided (or validated) by a human and then tracked robustly through vision predicting it with the UAV odometry.
With these results, discussed in [6], it was apparent that a new solution was needed, as the LiDAR approaches were too slow and the vision-based techniques probed themselves unreliable. The final proposed solution was based on integrating data from the laser and the vision sensors: the RANSAC over LiDAR approach would detect robustly the pipe and provide an initial position, which would then be projected into the image space (accounting for displacements if odometry is available) and used as a seed for the vision-based pipeline described.
In that same work [6], a sensibility analysis studying the effects of the relative pose between the sensor and pipes is provided. Once the pipe is detected in the LiDAR’s space sensor, the cylinder model is projected into the R2 image space using a projection matrix derived from the calibrated camera model (assumed to be a thin lens pinhole model, per classic literature [13]). This provides a region or band of interest where to look for the edges of the pipe in the image and is useful to solve the degenerate conic equation up to scale (i.e., being a function of the radius). An updated architecture version of the process is depicted in Figure 2.
The architecture of the multimodal perception pipeline combining LiDAR and camera vision. An updated version adds to previous works a validation step using odometric measurements.
The detailed architecture of the multimodal approach reveals how the LiDAR-based pipeline minimizes the data dimensionality by filtering non-curved surfaces (i.e., remove walls, floor, etc.) and also by removing entirely regions of the sensed space if priors or relevant data or the expected relative position of the pipe to the sensor is available. This was aimed at minimizing the size of the point cloud to be processed by the RANSAC step. To be able to project the detected pipe from the LiDAR sensor space into the camera image, some additional information was required: the rigid transformation between sensors (i.e., the calibration between LiDAR and camera) and an estimation of the odometry of the UAV. This is due because, even in the best assumption, with a performance slightly over 4 Hz, the delay between the captured point cloud and the produced estimation of the pipe would be over 200 ms. Therefore, the projection of the detected pipe to predict the area of interest to search the apparent contour has to consider the displacement during this period, not only the rigid LiDAR to camera transformation. This predicted region of interest is used in the vision process pipeline, with predictions of the appearance of the pipes into image space used to refine the contour search. This contour search relies on stacking a Hough transform to join line segment detector (LSD) detected segments (to overcome partial obstructions) on the relevant area and allows to choose the nearest correctly aligned lines. Notice that using a visual servoing library [14], an option to use data provided through human interaction was kept as available, though the integration of LiDAR detections as seeds into the visual pipeline made it unnecessary. To avoid degenerate or spurious solutions, a validation step (based on reprojection and “matching” of the Plückerian coordinates [15] for a tracked piped) was later introduced.
This architecture leads to a fast (limited by the performance of the vision-based part) and robust (based on the RANSAC resilience to spurious detections) pipe detector with great accuracy, which was deployed and test in a UAV. The main issue of the approach is the hardware requirements: access to odometry from the avionics systems, LiDAR, and camera sensors, and enough computing power to process them (beyond any other task required from the UAV). All this hardware is focused on solving what can be described as a semantic segmentation problem. This is relevant given the enormous changes produced in the last decade in the computer vision field, and how classic problems like semantic segmentation are currently solved.
In the context of computer vision, the semantic segmentation problem is used to determine which regions of an image present an object of a given category, that is, a class or label is assigned to a given area (be it a pixel, window, or segmented region). The different granularity accepted is produced by how the technique and its solution evolved: for a long time, it was completely unfeasible to produce pixel-wise solutions, so images were split according to different procedures, which added a complexity layer to the problem.
Current off-the-shelf technologies have changed the paradigm, as GPUs present huge capabilities in terms of parallelization, while solid-state disks make fast reliable storage cheap. These technical advancements have increased dramatically the performance, complexity, and memory available for data representation, especially for techniques inherently strong in highly parallelized environments. One of the fields where the impact has been more noticeable has been the artificial intelligence community, where the artificial neural network (ANN) has seen a resurgence thanks to the support this kind of hardware provides to otherwise computationally unfeasible techniques. The most impactful development in recent years has been the convolutional neural networks (CNNs), which have become the most popular computer vision approach for several of the classic problem and the default solution for semantic segmentation.
To understand the impact of deep learning into our proposed solution, we will discuss briefly how the classical segmentation pipeline worked and how the modern CNN-based classifier became the modern semantic segmentation techniques.
The classic semantic segmentation pipeline can be split into two generic blocks, namely image processing for feature extraction and feature level classification. The first block generally includes any image preprocessing done and resizing/resampling, splitting the image into the regions/windows, defining the granularity level of the classification, and finally, extracting the features itself. The features can be of any type and frequently the ones feed to the classification modules will be a composition of several individual features from different detectors. The use of different window/region-based approaches helps build up higher level features, and the classification can be refined at later stages with data from adjacent regions.
Notice that this kind of architecture generally relies on classifiers which required very accurate knowledge or a dataset with the classes to learn specified for each input so it can be trained. Figure 3 shows the detection of pipelines in classic semantic segmentation. Notice that to train the classifier, the image mask or classification result becomes also an input for the training process.
Block diagram of a classical architecture approach for semantic segmentation using computer vision.
So, it can be seen that solving the semantic segmentation problem though classic pattern recognition methods requires acute insight into the specifics of the problem domain, as the features to be detected and extracted are built/designed specifically. This implies (as mentioned earlier) working from low-level features and explicitly deriving the higher level features from them is a very complex problem itself, as they are affected by the input characteristics, what is to be found/discriminated, and which techniques will be used in the classification part of the pipeline.
Modern semantic segmentation techniques have organically evolved with the rise of the deep learning field to its current prominence. This evolution can be seen as a refinement in the scale of the inference produced from very coarse (image level probabilistic detection) to very fine (pixel level classification). The earliest ANN examples made probabilistic predictions about the presence of an object of a given class, that is, detection of objects with a probability assigned. The next step, achieved thanks to increased parallelization and network depth, was starting to tackle the localization problem, providing centroids and/or boxes for the detected classes (the use of classes instead of objects here is deliberate, as the instance segmentation problem, separating adjacent objects of the same class, would be dealt with much later).
The first big break into the classification problem was done by AlexNet [13] in 2012, when it won the ILSVRC challenge, with a score of 84.6% in the top-5 accuracy test, while the next best score was only 73.8% (based on classic techniques). AlexNet has since then become a known standard and a default network architecture to test problems, as it is actually not very deep or complex (see Figure 4). It presents five convolutional layers, with max-pooling after the first two, three fully connected layers, and a ReLU to deal with non-linearities. This clear victory of the CNN-based approaches was validated next year by Oxford’s VGG16 [16], one of the several architectures presented, winning the ILSVRC challenge with a 92.7% score.
Diagram of the AlexNet architecture, showcasing its pioneering use of convolutional layers.
While several other networks have been presented with deeper architecture, relevant development focused on introducing new types of structures into the networks. GoogLeNet [17], the 2014 ILSVRC winner, achieved victory thanks to the novel contribution of the inception module, which validated the concept that the CNN layers of a network could operate in other orders different from the classic sequential approach. Another relevant contribution produced by technology giants was ResNet [18], which scored a win for Microsoft in 2016. The introduction of residual blocks allowed them to increase the depth to 152 layers while keeping initial data meaningful for training the deeper layers. These residual blocks architecture essentially forwards a copy of the received inputs of a layer; thus, later layers received the results and same inputs of prior layers and can learn from the residuals.
More recently, ReNet [19] architecture was used to extend recurrent neural networks (RNNs) to multidimensional inputs.
The jump from the classification problem with some spatial data to pixel level labeling (refining inference from image/region to pixel level) was presented by Long [20], with the fully convolutional network (FCN). The method they proposed was based on using the full classifier (like the ones just discussed) as layers in a convolutional network architecture. FCN architecture, and its derivatives like U-Net [21] are the best solutions to semantic segmentation for most domains. These derivatives may include classic methods, such as DeepLab’s [22] conditional random fields [23], which reinforces the inference from spatially distant dependencies, usually lost due to CNN spatial invariance. The latest promising contributions to the semantic segmentation problem are based on the encoder-decoder architecture, known as autoenconders, like for example SegNet [24].
For the works discussed in this chapter, a FCN16 model with AlexNet as a semantic segmentation model was used. The main innovation introduced by the general FCN was exploiting the classification power via convolution of the common semantic segmentation DL network, but at the same time, reversing the downsampling effect of the convolution operation itself. Taking AlexNet as an example, as seen in Figure 4, convolutional layers apply a filter like operation while reducing the size of the data forwarded to the next layer. This process allows producing more accurate “deep features” but at the same time also removes high-level information describing the spatial relation between the features found. Thus, in order to exploit the features from the deep layers while the keeping information from spatial relation, data from multiple layers has to be fused (with element-wise summation). In order to be able to produce this fusion, data from the deeper layers are upsampled using deconvolution. Notice that data from shallow layers will be coarser but contain more spatial information. Thus, up to three different levels can be processed through FCN, depending on the quantity of layers deconvoluted and fused, as seen in Figure 5.
Detail of the skip architectures (FCN32, FCN16, and FCN8) used to produce results with data from several layers to recover both deep features and spatial information from shallow layers (courtesy of [25]).
More information on the detailed working of the different FCN models can be found in [25]. It is still worth noting that the more shallow layers are fused, the more accurate the model becomes, but according to the literature, the gain from FCN16 to FCN8 is minimal (below 2%).
Classic methods using trained classifiers would pick designed features (based on several metrics and detectors, as discussed earlier) to parametrize a given sample and assign a label. This would allow creating small specific datasets, which could be used to infer the knowledge to create bigger datasets in a posterior step. The high specificity of the features chosen (generally with expert domain knowledge applied implicitly) with respect to the task generally made them unsuitable to export learning to other domains.
By contrast, deep learning offers several transfer learning options. That is, as it was proven by Yosinski [26], trained with a distant domain dataset are generally useful for different domains and usually better than training from an initial random state. Notice that the transferability of features decreases with the difference between the previously trained task and the target one and implies that the network architecture is the same up to the transferred layers at least.
With this concept in mind, we decided to build a dataset to train an outdoor industrial pipe detector with pixel level annotation to be able to determine the position of the pipe. While the ability of transfer learning allows us to skip building a dataset with several tens of thousands of images, and therefore, the authors will work with a few thousand, which were used to fine-tune the network. These orders of magnitude are required as a “shallow” deep network. For instance, the AlexNet already presents 60 million parameters.
Capturing and labeling a dataset is a cumbersome task, so we also set to automatize this task with minimal human supervision/interaction, exploiting the capabilities of the sensing architecture proposed in earlier works described in Section 2.
This framework, see Figure 6, uses the images captured by the UAV camera sensor, the data processed by the localization approach chosen (see Section 2) to obtain the UAV odometry, and pipe detection seeds from the RANSAC technique treating the LiDAR point cloud data. When a pipe (or generally a cylinder) is detected and segmented in the data sensor provided by the LiDAR, this is used to produce a label for the temporally near images, to identify the region of the image (the set of pixels) containing the pipe or cylinder detected and its pose w.r.t. the camera. Notice that even running the perception part, the camera works at a higher rate than the LiDAR, so the full odometric estimation is used to interpolate between pipe detections, to estimate where the label should be projected into the in-between images (just as it was described for the pipe prediction in Section 2).
The framework proposed to automatically produce labeled datasets with the multimodal perception UAV.
This methodology was used to create an initial labeled dataset with actual data captured in real industrial scenarios during test and development flights, as it will be discussed in the next section.
To evaluate the viability of the proposed automated dataset generation methodology, we apply it to capture a dataset and train several semantic segmentation networks with it. To provide some quantitative quality measurement for the solutions produced, we use modified standard metrics for state-of-the-art deep learning, accounting that in our problem we are dealing with only one semantic class:
PA (pixel accuracy): base metric, defined by the ratio between properly classified pixels TP and the total number of pixels in an image, pixtotal:
Notice that usually, besides the PA the mean pixel accuracy (MPA) is provided, but in our case, it reduces to the same value of PA, thus it will not be provided.
IoU (intersection over union): standard metric in segmentation. The ration is computed between the intersection and union of two sets, namely the found segmentation and the labeled ground truth. Conceptually, it equals to the ratio between the number of correct positives (i.e., the intersection of the sets) TP, over all the correct positives, spurious positives FP and false negatives FN (i.e., the union of both the ground truth and segmentation provided). Usually, it is used as mean IoU (MIU), averaging the same ratio for all classes.
An additional metric usually computed along with the MIU is the frequency weighted MIU, which just weighs the average IoU computed at MIU according to the relative frequency of each class. The MIU, in our case, IoU is the most relevant metric and the most widely used when reporting segmentation results (semantic or otherwise).
The system proposed was implemented over the ROS meta-operating system, just as in previous works [6], where the UAV system used to capture the data is described. A set of real flights in simulated industry environments was performed, where flights around a pipe were done. During these flights, averaging ~240 s, an OTS USB camera was used to capture images (at 640 × 480 resolution), achieving an average frame rate of around 17 fps. This translated in around 20,000 raw images captured, including the parts of flight where no industry-like elements are present, thus of limited use.
Notice that as per the method described, the pipe to be found can be only labeled automatically when the LiDAR sensor can detect it; thus, the number of images was further reduced due to the range limitations of the LiDAR scanner. Other factors, such as vibrations and disruptions in the input or results of required perceptual data, further reduced the number of images with accurate labels.
Around ~2100 images were automatically labeled with a mask assigning a ground truth for the pipe in the image. After an initial human inspection of the assigned label, a further ~320 were rejected, obtaining a final set of 1750. The image rejected produced spurious ground truths/masks. Some of them had inconsistent data and the reprojection of the cylinder detected in through RANSAC in LiDAR scans was not properly aligned (error could be produced by spurious interpolation of poses, faulty synchronization data from the sensors, or due to deformation of the UAV frame, as it is impossible for it to be perfectly rigid). Another group presented partial detections (only one of the edges of the pipe is visible in the image), thus making it useless for the apparent contour optimization. A third type of error found was produced by the vision-based pipeline, where a spurious mask was generated, commonly some shadows/textures displace/retort the edge, or areas not pertaining to the pipe are assigned due similarity of the texture and complexity of delimiting the areas.
A sample of the labeling process can be seen in Figure 7, with the original image, the segmented pipe image, and approximations to centroid and bounding bow.
Left: dataset image. Middle: bounding box and centroid of the region detected. Right: segmentation mask image.
Out of the several options available to test the validity of the dataset produced, the shallow architecture AlexNet was selected, as it could be easily trained and it would provide some insight in the performance that could be realistically expected from a CNN-based approach deployed in the limited hardware of a UAV.
According to previous literature, the dataset was divided into training, validation, and test at the standard ratio of 70, 15, and 15% respectively.
To match the input of AlexNet the images were resized to 256 × 256 resolution. This was mainly done to reduce the computational load, as the input size could be easily fit adjusting some parameters, like the stride. To train and test the network, the Pytorch library was used, which provides full support for its own implementation of AlexNet.
To produce some metrics relevant to the network architecture just trained, a modified version of the technique used to label the dataset was used. Note that this approach, as described in previous sections, uses LiDAR, cameras, and odometry to: acquire an initial robust detection (from LiDAR), track its projection and predict it in the camera image space (using odometric data), and finally determine its edges/contour in the image. The robustness of the LiDAR detection is mainly due to exploiting prior knowledge (in the form of the known radius of the pipe to detect) that cannot be introduced into the AlexNet architecture to produce a meaningful comparison. So, a modified method, referred to as NPMD (no-priors multimodal detector) was employed to estimate the accuracy of earlier work detector without priors. The main difference was modifying the LiDAR pipeline to be able to detect several pipes with different radius (as it should be considered unknown). This led to the appearance of false positives and spurious measurements, which in turn weakened the results produced by the segmentation part of the visual pipeline.
Thus, FCN with AlexNet classification was trained using a pre-trained model for AlexNet, with the standard stochastic gradient descend (SGD) with a momentum of 0.9. A learning rate of 10−3 was used, according to known literature, with image batches of 20. The weight decay and bias learning rate were set to standard values of 5.10−4 and 2, respectively. Without any prior data, and no benefit to obtain by doing otherwise reported in any previous works, the classifier layer was set to 0, and the dropout layer in the AlexNet left unmodified. This trained model produced the results found in Table 1.
AlexNetFCN | UPMD | |
---|---|---|
PA | 73.4 | 56.7 |
IoU | 58.6 | 42.1 |
Experimental results obtained by AlexNet-based FCN.
It can be seen that eliminating the seed/prior data from the multimodal detector made it rather weak, with very low values for IoU, signaling the presence of spurious detections and probably fake positives. The FCN-based solution was around 1.5 times better segmenting the pipe, being a clear winner. This was to be expected as we deliberately removed one of the key factors contributing to the LiDAR-based RANSAC detection robustness, the radius priors, leading to the appearance of spurious detections.
It is worth noting that although the results are not that strong in terms of metrics achieved for a single-class case, there are no other vision-only pipe detectors with better results in the literature, neither other approaches actually tested in real UAV’s platforms, like authors’ previous works [6].
The field of computer vision has been greatly impacted by the advances in deep learning that have emerged in the last decade. This has allowed solving, with purely vision-based approaches, some problems that were considered unsolvable under this restriction. In the case presented, a detection and positioning problem, solved with limited hardware resources (onboard a UAV) in an industry-like uncontrolled scenario through a multimodal approach, has been solved with a vision-only approach. The previous multimodal approach relied in LiDAR, cameras, and odometric measurements (mainly from GPS and IMU) to extract data with complex algorithms like RANSAC and combine them to predict the position of a pipe and produce a measurement. This system was notable thanks to its robustness and performance but presented the huge requirements detailed in [6]. In order to solve the problem in a simpler and more affordable manner, a pure visual solution was chosen as the way to go, exploring the deep learning opportunities.
Although the switch to a pure visual solution meant that during its use, the procedure would only use the camera sensor, the multimodal approach was still used to capture data, and through a series of modifications, turn it into an automatic labeling tool. This allowed building a small but complete dataset with fully labeled images relevant to the problem that we were trying to solve. Finally, to test this dataset, we train a DL architecture able to solve the semantic segmentation problem. Thus, three different contributions have been discussed in this chapter: firstly, a dataset generator exploiting multimodal data captured by the perception system to be replaced has been designed and implemented; secondly, with this dataset generation tool, the data captured has been properly labeled so it can be used for DL applications; and finally, a sample lightweight network model for semantic segmentation, FCN with AlexNet classification, has been trained and evaluated to test the problem.
By the same reasons that there was no dataset available for our challenge and we had to capture and develop one dedicated to our domain, there were no related works to obtain metrics. In order to have some relevant metrics to compare the results of the developed approach, a modified version of [13] was produced and benchmarked without the use of prior knowledge. Under these assumptions, the new CNN-based method was able to clearly surpass the multimodal approach, though it still lacks robustness to be considered ready for industrial standards. Still, these initial tests have proven the viability of the built dataset generator and the utilization of CNN-based semantic segmentation to replace the multimodal approach.
This research was funded by the Spanish Ministry of Economy, Industry and Competitiveness through Project 2016-78957-R.
Authors are listed below with their open access chapters linked via author name:
",metaTitle:"IntechOpen authors on the Global Highly Cited Researchers 2018 list",metaDescription:null,metaKeywords:null,canonicalURL:null,contentRaw:'[{"type":"htmlEditorComponent","content":"New for 2018 (alphabetically by surname).
\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nJocelyn Chanussot (chapter to be published soon...)
\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nYuekun Lai
\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nPrevious years (alphabetically by surname)
\\n\\nAbdul Latif Ahmad 2016-18
\\n\\nKhalil Amine 2017, 2018
\\n\\nEwan Birney 2015-18
\\n\\nFrede Blaabjerg 2015-18
\\n\\nGang Chen 2016-18
\\n\\nJunhong Chen 2017, 2018
\\n\\nZhigang Chen 2016, 2018
\\n\\nMyung-Haing Cho 2016, 2018
\\n\\nMark Connors 2015-18
\\n\\nCyrus Cooper 2017, 2018
\\n\\nLiming Dai 2015-18
\\n\\nWeihua Deng 2017, 2018
\\n\\nVincenzo Fogliano 2017, 2018
\\n\\nRon de Graaf 2014-18
\\n\\nHarald Haas 2017, 2018
\\n\\nFrancisco Herrera 2017, 2018
\\n\\nJaakko Kangasjärvi 2015-18
\\n\\nHamid Reza Karimi 2016-18
\\n\\nJunji Kido 2014-18
\\n\\nJose Luiszamorano 2015-18
\\n\\nYiqi Luo 2016-18
\\n\\nJoachim Maier 2014-18
\\n\\nAndrea Natale 2017, 2018
\\n\\nAlberto Mantovani 2014-18
\\n\\nMarjan Mernik 2017, 2018
\\n\\nSandra Orchard 2014, 2016-18
\\n\\nMohamed Oukka 2016-18
\\n\\nBiswajeet Pradhan 2016-18
\\n\\nDirk Raes 2017, 2018
\\n\\nUlrike Ravens-Sieberer 2016-18
\\n\\nYexiang Tong 2017, 2018
\\n\\nJim Van Os 2015-18
\\n\\nLong Wang 2017, 2018
\\n\\nFei Wei 2016-18
\\n\\nIoannis Xenarios 2017, 2018
\\n\\nQi Xie 2016-18
\\n\\nXin-She Yang 2017, 2018
\\n\\nYulong Yin 2015, 2017, 2018
\\n"}]'},components:[{type:"htmlEditorComponent",content:'New for 2018 (alphabetically by surname).
\n\n\n\n\n\n\n\n\n\nJocelyn Chanussot (chapter to be published soon...)
\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nYuekun Lai
\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nPrevious years (alphabetically by surname)
\n\nAbdul Latif Ahmad 2016-18
\n\nKhalil Amine 2017, 2018
\n\nEwan Birney 2015-18
\n\nFrede Blaabjerg 2015-18
\n\nGang Chen 2016-18
\n\nJunhong Chen 2017, 2018
\n\nZhigang Chen 2016, 2018
\n\nMyung-Haing Cho 2016, 2018
\n\nMark Connors 2015-18
\n\nCyrus Cooper 2017, 2018
\n\nLiming Dai 2015-18
\n\nWeihua Deng 2017, 2018
\n\nVincenzo Fogliano 2017, 2018
\n\nRon de Graaf 2014-18
\n\nHarald Haas 2017, 2018
\n\nFrancisco Herrera 2017, 2018
\n\nJaakko Kangasjärvi 2015-18
\n\nHamid Reza Karimi 2016-18
\n\nJunji Kido 2014-18
\n\nJose Luiszamorano 2015-18
\n\nYiqi Luo 2016-18
\n\nJoachim Maier 2014-18
\n\nAndrea Natale 2017, 2018
\n\nAlberto Mantovani 2014-18
\n\nMarjan Mernik 2017, 2018
\n\nSandra Orchard 2014, 2016-18
\n\nMohamed Oukka 2016-18
\n\nBiswajeet Pradhan 2016-18
\n\nDirk Raes 2017, 2018
\n\nUlrike Ravens-Sieberer 2016-18
\n\nYexiang Tong 2017, 2018
\n\nJim Van Os 2015-18
\n\nLong Wang 2017, 2018
\n\nFei Wei 2016-18
\n\nIoannis Xenarios 2017, 2018
\n\nQi Xie 2016-18
\n\nXin-She Yang 2017, 2018
\n\nYulong Yin 2015, 2017, 2018
\n'}]},successStories:{items:[]},authorsAndEditors:{filterParams:{sort:"featured,name"},profiles:[{id:"289905",title:"Dr.",name:null,middleName:null,surname:"Inamuddin",slug:"inamuddin",fullName:"Inamuddin",position:null,profilePictureURL:"https://mts.intechopen.com/storage/users/289905/images/system/289905.jpeg",biography:"Dr. Inamuddin is currently working as an assistant professor in the Chemistry Department, Faculty of Science, King Abdulaziz University, Jeddah, Saudi Arabia. He has extensive research experience in multidisciplinary fields of analytical chemistry, materials chemistry, electrochemistry, and more specifically, renewable energy and the environment. He has published 127 research articles in international journals of repute and 18 book chapters in knowledge-based book editions published by renowned international publishers. He has published 39 edited books with Springer, United Kingdom, Elsevier, Nova Science Publishers, Inc. USA, CRC Press Taylor & Francis, Asia Pacific, Trans Tech Publications Ltd., Switzerland, and Materials Science Forum, USA. He is a member of various editorial boards serving as associate editor for journals such as Environmental Chemistry Letter, Applied Water Science, Euro-Mediterranean Journal for Environmental Integration, Springer-Nature, Scientific Reports-Nature, and the editor of Eurasian Journal of Analytical Chemistry.",institutionString:"King Abdulaziz University",institution:{name:"King Abdulaziz University",country:{name:"Saudi Arabia"}}},{id:"99002",title:"Dr.",name:null,middleName:null,surname:"Koontongkaew",slug:"koontongkaew",fullName:"Koontongkaew",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:null,institutionString:null,institution:{name:"Thammasat University",country:{name:"Thailand"}}},{id:"156647",title:"Dr.",name:"A K M Mamunur",middleName:null,surname:"Rashid",slug:"a-k-m-mamunur-rashid",fullName:"A K M Mamunur Rashid",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:"MBBS, DCH, MD(Paed.), Grad. Cert. P. Rheum.(UWA, Australia), FRCP(Edin.)",institutionString:null,institution:{name:"Khulna Medical College",country:{name:"Bangladesh"}}},{id:"234696",title:"Prof.",name:"A K M Mominul",middleName:null,surname:"Islam",slug:"a-k-m-mominul-islam",fullName:"A K M Mominul Islam",position:null,profilePictureURL:"https://intech-files.s3.amazonaws.com/a043Y00000cA8dpQAC/Co2_Profile_Picture-1588761796759",biography:"Prof. Dr. A. K. M. Mominul Islam received both of his bachelor's and Master’s degree from Bangladesh Agricultural University. After that, he joined as Lecturer of Agronomy at Bangladesh Agricultural University (BAU), Mymensingh, Bangladesh, and became Professor in the same department of the university. Dr. Islam did his second Master’s in Physical Land Resources from Ghent University, Belgium. He is currently serving as a postdoctoral researcher at the Department of Horticulture & Landscape Architecture at Purdue University, USA. Dr. Islam has obtained his Ph.D. degree in Plant Allelopathy from The United Graduate School of Agricultural Sciences, Ehime University, Japan. The dissertation title of Dr. Islam was “Allelopathy of five Lamiaceae medicinal plant species”. Dr. Islam is the author of 38 articles published in nationally and internationally reputed journals, 1 book chapter, and 3 books. He is a member of the editorial board and referee of several national and international journals. He is supervising the research of MS and Ph.D. students in areas of Agronomy. Prof. Islam is conducting research on crop management, bio-herbicides, and allelopathy.",institutionString:"Bangladesh Agricultural University",institution:{name:"Bangladesh Agricultural University",country:{name:"Bangladesh"}}},{id:"214531",title:"Mr.",name:"A T M Sakiur",middleName:null,surname:"Rahman",slug:"a-t-m-sakiur-rahman",fullName:"A T M Sakiur Rahman",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:null,institutionString:null,institution:{name:"University of Rajshahi",country:{name:"Bangladesh"}}},{id:"66545",title:"Dr.",name:"A. F.",middleName:null,surname:"Omar",slug:"a.-f.-omar",fullName:"A. F. Omar",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:"Dr. A. F. Omar obtained\nhis Bachelor degree in electrical and\nelectronics engineering from Universiti\nSains Malaysia in 2002, Master of Science in electronics\nengineering from Open University\nMalaysia in 2008 and PhD in optical physics from Universiti\nSains Malaysia in 2012. His research mainly\nfocuses on the development of optical\nand electronics systems for spectroscopy\napplication in environmental monitoring,\nagriculture and dermatology. He has\nmore than 10 years of teaching\nexperience in subjects related to\nelectronics, mathematics and applied optics for\nuniversity students and industrial engineers.",institutionString:null,institution:{name:"Universiti Sains Malaysia",country:{name:"Malaysia"}}},{id:"191072",title:"Prof.",name:"A. K. M. Aminul",middleName:null,surname:"Islam",slug:"a.-k.-m.-aminul-islam",fullName:"A. K. M. Aminul Islam",position:null,profilePictureURL:"https://mts.intechopen.com/storage/users/191072/images/system/191072.jpg",biography:"Prof. Dr. A. K. M. Aminul Islam received both of his bachelor and Master’s degree from Bangladesh Agricultural University. After that he joined as Lecturer of Genetics and Plant Breeding at Bangabandhu Sheikh Mujibur Rahman Agricultural University (BSMRAU), Gazipur, Bangladesh and became Professor in the same department of the university. He is currently serving as Director (Research) of Bangabandhu Sheikh Mujibur Rahman Agricultural University (BSMRAU), Gazipur, Bangladesh. Dr. Islam has obtained his Ph D degree in Chemical and Process Engineering from Universiti Kebangsaan Malaysia. The dissertation title of Dr. Islam was “Improvement of Biodiesel Production through Genetic Studies of Jatropha (Jatropha curcas L.)”. Dr. Islam is the author of 98 articles published in nationally and internationally reputed journals, 11 book chapters and 3 books. He is a member of editorial board and referee of several national and international journals. He is also serving as the General Secretary of Plant Breeding and Genetics Society of Bangladesh, Seminar and research Secretary of JICA Alumni Association of Bangladesh and member of several professional societies. Prof. Islam acted as Principal Breeder in the releasing system of BU Hybrid Lau 1, BU Lau 1, BU Capsicum 1, BU Lalshak 1, BU Baromashi Seem 1, BU Sheem 1, BU Sheem 2, BU Sheem 3 and BU Sheem 4. He supervised 50 MS and 3 Ph D students. Prof. Islam currently supervising research of 5 MS and 3 Ph D students in areas Plant Breeding & Seed Technologies. Conducting research on development of hybrid vegetables, hybrid Brassica napus using CMS system, renewable energy research with Jatropha curcas.",institutionString:"Bangabandhu Sheikh Mujibur Rahman Agricultural University",institution:{name:"Bangabandhu Sheikh Mujibur Rahman Agricultural University",country:{name:"Bangladesh"}}},{id:"322225",title:"Dr.",name:"A. K. M. Aminul",middleName:null,surname:"Islam",slug:"a.-k.-m.-aminul-islam",fullName:"A. K. M. Aminul Islam",position:null,profilePictureURL:"https://mts.intechopen.com/storage/users/no_image.jpg",biography:"Prof. Dr. A. K. M. Aminul Islam received both of his bachelor's and Master’s degree from Bangladesh Agricultural University. After that he joined as Lecturer of Genetics and Plant Breeding at Bangabandhu Sheikh Mujibur Rahman Agricultural University (BSMRAU), Gazipur, Bangladesh, and became Professor in the same department of the university. He is currently serving as Director (Research) of Bangabandhu Sheikh Mujibur Rahman Agricultural University (BSMRAU), Gazipur, Bangladesh. Dr. Islam has obtained his Ph.D. degree in Chemical and Process Engineering from Universiti Kebangsaan Malaysia. The dissertation title of Dr. Islam was 'Improvement of Biodiesel Production through Genetic Studies of Jatropha (Jatropha curcas L.)”. Dr. Islam is the author of 99 articles published in nationally and internationally reputed journals, 11 book chapters, 3 books, and 20 proceedings and conference paper. He is a member of the editorial board and referee of several national and international journals. He is also serving as the General Secretary of Plant Breeding and Genetics Society of Bangladesh, Seminar, and research Secretary of JICA Alumni Association of Bangladesh and a member of several professional societies. Prof. Islam acted as Principal Breeder in the releasing system of BU Hybrid Lau 1, BU Lau 1, BU Capsicum 1, BU Lalshak 1, BU Baromashi Seem 1, BU Sheem 1, BU Sheem 2, BU Sheem 3 and BU Sheem 4. He supervised 50 MS and 3 PhD students. Prof. Islam currently supervising the research of 5 MS and 3 PhD students in areas Plant Breeding & Seed Technologies. Conducting research on the development of hybrid vegetables, hybrid Brassica napus using CMS system, renewable energy research with Jatropha curcas.",institutionString:"Bangabandhu Sheikh Mujibur Rahman Agricultural University",institution:{name:"Bangabandhu Sheikh Mujibur Rahman Agricultural University",country:{name:"Bangladesh"}}},{id:"91977",title:"Dr.",name:"A.B.M. Sharif",middleName:null,surname:"Hossain",slug:"a.b.m.-sharif-hossain",fullName:"A.B.M. Sharif Hossain",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:null,institutionString:null,institution:{name:"University of Malaya",country:{name:"Malaysia"}}},{id:"97123",title:"Prof.",name:"A.M.M.",middleName:null,surname:"Sharif Ullah",slug:"a.m.m.-sharif-ullah",fullName:"A.M.M. Sharif Ullah",position:null,profilePictureURL:"https://mts.intechopen.com/storage/users/97123/images/4209_n.jpg",biography:"AMM Sharif Ullah is currently an Associate Professor of Design and Manufacturing in Department of Mechanical Engineering at Kitami Institute of Technology, Japan. He received the Bachelor of Science Degree in Mechanical Engineering in 1992 from the Bangladesh University of Engineering and Technology, Dhaka, Bangladesh. In 1993, he moved to Japan for graduate studies. He received the Master of Engineering degree in 1996 from the Kansai University Graduate School of Engineering in Mechanical Engineering (Major: Manufacturing Engineering). He also received the Doctor of Engineering degree from the same institute in the same field in 1999. He began his academic career in 2000 as an Assistant Professor in the Industrial Systems Engineering Program at the Asian Institute of Technology, Thailand, as an Assistant Professor in the Industrial Systems Engineering Program. In 2002, he took up the position of Assistant Professor in the Department of Mechanical Engineering at the United Arab Emirates (UAE) University. He was promoted to Associate Professor in 2006 at the UAE University. He moved to his current employer in 2009. His research field is product realization engineering (design, manufacturing, operations, and sustainability). He teaches design and manufacturing related courses at undergraduate and graduate degree programs. He has been mentoring a large number of students for their senior design projects and theses. He has published more than 90 papers in refereed journals, edited books, and international conference proceedings. He made more than 35 oral presentations. Since 2005, he directs the advanced manufacturing engineering research laboratory at Kitami Institute of Technology.",institutionString:null,institution:{name:"Kitami Institute of Technology",country:{name:"Japan"}}},{id:"213441",title:"Dr.",name:"A.R.Kavitha",middleName:null,surname:"Balaji",slug:"a.r.kavitha-balaji",fullName:"A.R.Kavitha Balaji",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:null,institutionString:null,institution:{name:"Anna University, Chennai",country:{name:"India"}}},{id:"172688",title:"Prof.",name:"A.V.",middleName:null,surname:"Salker",slug:"a.v.-salker",fullName:"A.V. Salker",position:null,profilePictureURL:"//cdnintech.com/web/frontend/www/assets/author.svg",biography:null,institutionString:null,institution:{name:"Goa University",country:{name:"India"}}}],filtersByRegion:[{group:"region",caption:"North America",value:1,count:5684},{group:"region",caption:"Middle and South America",value:2,count:5166},{group:"region",caption:"Africa",value:3,count:1682},{group:"region",caption:"Asia",value:4,count:10211},{group:"region",caption:"Australia and Oceania",value:5,count:887},{group:"region",caption:"Europe",value:6,count:15616}],offset:12,limit:12,total:10241},chapterEmbeded:{data:{}},editorApplication:{success:null,errors:{}},ofsBooks:{filterParams:{topicId:"15"},books:[{type:"book",id:"10471",title:"Advances in Dynamical Systems Theory, Models, Algorithms and Applications",subtitle:null,isOpenForSubmission:!0,hash:"689fdf3cdc78ade03f0c43a245dcf818",slug:null,bookSignature:"Dr. Bruno Carpentieri",coverURL:"https://cdn.intechopen.com/books/images_new/10471.jpg",editedByType:null,editors:[{id:"92921",title:"Dr.",name:"Bruno",surname:"Carpentieri",slug:"bruno-carpentieri",fullName:"Bruno Carpentieri"}],productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"10677",title:"Topology",subtitle:null,isOpenForSubmission:!0,hash:"85eac84b173d785f989522397616124e",slug:null,bookSignature:"",coverURL:"https://cdn.intechopen.com/books/images_new/10677.jpg",editedByType:null,editors:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"10678",title:"Biostatistics",subtitle:null,isOpenForSubmission:!0,hash:"f63db439474a574454a66894db8b394c",slug:null,bookSignature:"",coverURL:"https://cdn.intechopen.com/books/images_new/10678.jpg",editedByType:null,editors:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}}],filtersByTopic:[{group:"topic",caption:"Agricultural and Biological Sciences",value:5,count:9},{group:"topic",caption:"Biochemistry, Genetics and Molecular Biology",value:6,count:18},{group:"topic",caption:"Business, Management and Economics",value:7,count:2},{group:"topic",caption:"Chemistry",value:8,count:7},{group:"topic",caption:"Computer and Information Science",value:9,count:10},{group:"topic",caption:"Earth and Planetary Sciences",value:10,count:5},{group:"topic",caption:"Engineering",value:11,count:15},{group:"topic",caption:"Environmental Sciences",value:12,count:2},{group:"topic",caption:"Immunology and Microbiology",value:13,count:5},{group:"topic",caption:"Materials Science",value:14,count:4},{group:"topic",caption:"Mathematics",value:15,count:1},{group:"topic",caption:"Medicine",value:16,count:60},{group:"topic",caption:"Nanotechnology and Nanomaterials",value:17,count:1},{group:"topic",caption:"Neuroscience",value:18,count:1},{group:"topic",caption:"Pharmacology, Toxicology and Pharmaceutical Science",value:19,count:6},{group:"topic",caption:"Physics",value:20,count:2},{group:"topic",caption:"Psychology",value:21,count:3},{group:"topic",caption:"Robotics",value:22,count:1},{group:"topic",caption:"Social Sciences",value:23,count:3},{group:"topic",caption:"Technology",value:24,count:1},{group:"topic",caption:"Veterinary Medicine and Science",value:25,count:2}],offset:12,limit:12,total:3},popularBooks:{featuredBooks:[{type:"book",id:"9208",title:"Welding",subtitle:"Modern Topics",isOpenForSubmission:!1,hash:"7d6be076ccf3a3f8bd2ca52d86d4506b",slug:"welding-modern-topics",bookSignature:"Sadek Crisóstomo Absi Alfaro, Wojciech Borek and Błażej Tomiczek",coverURL:"https://cdn.intechopen.com/books/images_new/9208.jpg",editors:[{id:"65292",title:"Prof.",name:"Sadek Crisostomo Absi",middleName:"C. Absi",surname:"Alfaro",slug:"sadek-crisostomo-absi-alfaro",fullName:"Sadek Crisostomo Absi Alfaro"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9139",title:"Topics in Primary Care Medicine",subtitle:null,isOpenForSubmission:!1,hash:"ea774a4d4c1179da92a782e0ae9cde92",slug:"topics-in-primary-care-medicine",bookSignature:"Thomas F. Heston",coverURL:"https://cdn.intechopen.com/books/images_new/9139.jpg",editors:[{id:"217926",title:"Dr.",name:"Thomas F.",middleName:null,surname:"Heston",slug:"thomas-f.-heston",fullName:"Thomas F. Heston"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"8697",title:"Virtual Reality and Its Application in Education",subtitle:null,isOpenForSubmission:!1,hash:"ee01b5e387ba0062c6b0d1e9227bda05",slug:"virtual-reality-and-its-application-in-education",bookSignature:"Dragan Cvetković",coverURL:"https://cdn.intechopen.com/books/images_new/8697.jpg",editors:[{id:"101330",title:"Dr.",name:"Dragan",middleName:"Mladen",surname:"Cvetković",slug:"dragan-cvetkovic",fullName:"Dragan Cvetković"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9785",title:"Endometriosis",subtitle:null,isOpenForSubmission:!1,hash:"f457ca61f29cf7e8bc191732c50bb0ce",slug:"endometriosis",bookSignature:"Courtney Marsh",coverURL:"https://cdn.intechopen.com/books/images_new/9785.jpg",editors:[{id:"255491",title:"Dr.",name:"Courtney",middleName:null,surname:"Marsh",slug:"courtney-marsh",fullName:"Courtney Marsh"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9343",title:"Trace Metals in the Environment",subtitle:"New Approaches and Recent Advances",isOpenForSubmission:!1,hash:"ae07e345bc2ce1ebbda9f70c5cd12141",slug:"trace-metals-in-the-environment-new-approaches-and-recent-advances",bookSignature:"Mario Alfonso Murillo-Tovar, Hugo Saldarriaga-Noreña and Agnieszka Saeid",coverURL:"https://cdn.intechopen.com/books/images_new/9343.jpg",editors:[{id:"255959",title:"Dr.",name:"Mario Alfonso",middleName:null,surname:"Murillo-Tovar",slug:"mario-alfonso-murillo-tovar",fullName:"Mario Alfonso Murillo-Tovar"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"7831",title:"Sustainability in Urban Planning and Design",subtitle:null,isOpenForSubmission:!1,hash:"c924420492c8c2c9751e178d025f4066",slug:"sustainability-in-urban-planning-and-design",bookSignature:"Amjad Almusaed, Asaad Almssad and Linh Truong - Hong",coverURL:"https://cdn.intechopen.com/books/images_new/7831.jpg",editors:[{id:"110471",title:"Dr.",name:"Amjad",middleName:"Zaki",surname:"Almusaed",slug:"amjad-almusaed",fullName:"Amjad Almusaed"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"8468",title:"Sheep Farming",subtitle:"An Approach to Feed, Growth and Sanity",isOpenForSubmission:!1,hash:"838f08594850bc04aa14ec873ed1b96f",slug:"sheep-farming-an-approach-to-feed-growth-and-sanity",bookSignature:"António Monteiro",coverURL:"https://cdn.intechopen.com/books/images_new/8468.jpg",editors:[{id:"190314",title:"Prof.",name:"António",middleName:"Cardoso",surname:"Monteiro",slug:"antonio-monteiro",fullName:"António Monteiro"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"8816",title:"Financial Crises",subtitle:"A Selection of Readings",isOpenForSubmission:!1,hash:"6f2f49fb903656e4e54280c79fabd10c",slug:"financial-crises-a-selection-of-readings",bookSignature:"Stelios Markoulis",coverURL:"https://cdn.intechopen.com/books/images_new/8816.jpg",editors:[{id:"237863",title:"Dr.",name:"Stelios",middleName:null,surname:"Markoulis",slug:"stelios-markoulis",fullName:"Stelios Markoulis"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9376",title:"Contemporary Developments and Perspectives in International Health Security",subtitle:"Volume 1",isOpenForSubmission:!1,hash:"b9a00b84cd04aae458fb1d6c65795601",slug:"contemporary-developments-and-perspectives-in-international-health-security-volume-1",bookSignature:"Stanislaw P. Stawicki, Michael S. Firstenberg, Sagar C. Galwankar, Ricardo Izurieta and Thomas Papadimos",coverURL:"https://cdn.intechopen.com/books/images_new/9376.jpg",editors:[{id:"181694",title:"Dr.",name:"Stanislaw P.",middleName:null,surname:"Stawicki",slug:"stanislaw-p.-stawicki",fullName:"Stanislaw P. Stawicki"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"7769",title:"Medical Isotopes",subtitle:null,isOpenForSubmission:!1,hash:"f8d3c5a6c9a42398e56b4e82264753f7",slug:"medical-isotopes",bookSignature:"Syed Ali Raza Naqvi and Muhammad Babar Imrani",coverURL:"https://cdn.intechopen.com/books/images_new/7769.jpg",editors:[{id:"259190",title:"Dr.",name:"Syed Ali Raza",middleName:null,surname:"Naqvi",slug:"syed-ali-raza-naqvi",fullName:"Syed Ali Raza Naqvi"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9279",title:"Concepts, Applications and Emerging Opportunities in Industrial Engineering",subtitle:null,isOpenForSubmission:!1,hash:"9bfa87f9b627a5468b7c1e30b0eea07a",slug:"concepts-applications-and-emerging-opportunities-in-industrial-engineering",bookSignature:"Gary Moynihan",coverURL:"https://cdn.intechopen.com/books/images_new/9279.jpg",editors:[{id:"16974",title:"Dr.",name:"Gary",middleName:null,surname:"Moynihan",slug:"gary-moynihan",fullName:"Gary Moynihan"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"7807",title:"A Closer Look at Organizational Culture in Action",subtitle:null,isOpenForSubmission:!1,hash:"05c608b9271cc2bc711f4b28748b247b",slug:"a-closer-look-at-organizational-culture-in-action",bookSignature:"Süleyman Davut Göker",coverURL:"https://cdn.intechopen.com/books/images_new/7807.jpg",editors:[{id:"190035",title:"Associate Prof.",name:"Süleyman Davut",middleName:null,surname:"Göker",slug:"suleyman-davut-goker",fullName:"Süleyman Davut Göker"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}}],offset:12,limit:12,total:5131},hotBookTopics:{hotBooks:[],offset:0,limit:12,total:null},publish:{},publishingProposal:{success:null,errors:{}},books:{featuredBooks:[{type:"book",id:"9208",title:"Welding",subtitle:"Modern Topics",isOpenForSubmission:!1,hash:"7d6be076ccf3a3f8bd2ca52d86d4506b",slug:"welding-modern-topics",bookSignature:"Sadek Crisóstomo Absi Alfaro, Wojciech Borek and Błażej Tomiczek",coverURL:"https://cdn.intechopen.com/books/images_new/9208.jpg",editors:[{id:"65292",title:"Prof.",name:"Sadek Crisostomo Absi",middleName:"C. Absi",surname:"Alfaro",slug:"sadek-crisostomo-absi-alfaro",fullName:"Sadek Crisostomo Absi Alfaro"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9139",title:"Topics in Primary Care Medicine",subtitle:null,isOpenForSubmission:!1,hash:"ea774a4d4c1179da92a782e0ae9cde92",slug:"topics-in-primary-care-medicine",bookSignature:"Thomas F. Heston",coverURL:"https://cdn.intechopen.com/books/images_new/9139.jpg",editors:[{id:"217926",title:"Dr.",name:"Thomas F.",middleName:null,surname:"Heston",slug:"thomas-f.-heston",fullName:"Thomas F. Heston"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"8697",title:"Virtual Reality and Its Application in Education",subtitle:null,isOpenForSubmission:!1,hash:"ee01b5e387ba0062c6b0d1e9227bda05",slug:"virtual-reality-and-its-application-in-education",bookSignature:"Dragan Cvetković",coverURL:"https://cdn.intechopen.com/books/images_new/8697.jpg",editors:[{id:"101330",title:"Dr.",name:"Dragan",middleName:"Mladen",surname:"Cvetković",slug:"dragan-cvetkovic",fullName:"Dragan Cvetković"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9785",title:"Endometriosis",subtitle:null,isOpenForSubmission:!1,hash:"f457ca61f29cf7e8bc191732c50bb0ce",slug:"endometriosis",bookSignature:"Courtney Marsh",coverURL:"https://cdn.intechopen.com/books/images_new/9785.jpg",editors:[{id:"255491",title:"Dr.",name:"Courtney",middleName:null,surname:"Marsh",slug:"courtney-marsh",fullName:"Courtney Marsh"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9343",title:"Trace Metals in the Environment",subtitle:"New Approaches and Recent Advances",isOpenForSubmission:!1,hash:"ae07e345bc2ce1ebbda9f70c5cd12141",slug:"trace-metals-in-the-environment-new-approaches-and-recent-advances",bookSignature:"Mario Alfonso Murillo-Tovar, Hugo Saldarriaga-Noreña and Agnieszka Saeid",coverURL:"https://cdn.intechopen.com/books/images_new/9343.jpg",editors:[{id:"255959",title:"Dr.",name:"Mario Alfonso",middleName:null,surname:"Murillo-Tovar",slug:"mario-alfonso-murillo-tovar",fullName:"Mario Alfonso Murillo-Tovar"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"8468",title:"Sheep Farming",subtitle:"An Approach to Feed, Growth and Sanity",isOpenForSubmission:!1,hash:"838f08594850bc04aa14ec873ed1b96f",slug:"sheep-farming-an-approach-to-feed-growth-and-sanity",bookSignature:"António Monteiro",coverURL:"https://cdn.intechopen.com/books/images_new/8468.jpg",editors:[{id:"190314",title:"Prof.",name:"António",middleName:"Cardoso",surname:"Monteiro",slug:"antonio-monteiro",fullName:"António Monteiro"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"8816",title:"Financial Crises",subtitle:"A Selection of Readings",isOpenForSubmission:!1,hash:"6f2f49fb903656e4e54280c79fabd10c",slug:"financial-crises-a-selection-of-readings",bookSignature:"Stelios Markoulis",coverURL:"https://cdn.intechopen.com/books/images_new/8816.jpg",editors:[{id:"237863",title:"Dr.",name:"Stelios",middleName:null,surname:"Markoulis",slug:"stelios-markoulis",fullName:"Stelios Markoulis"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"7831",title:"Sustainability in Urban Planning and Design",subtitle:null,isOpenForSubmission:!1,hash:"c924420492c8c2c9751e178d025f4066",slug:"sustainability-in-urban-planning-and-design",bookSignature:"Amjad Almusaed, Asaad Almssad and Linh Truong - Hong",coverURL:"https://cdn.intechopen.com/books/images_new/7831.jpg",editors:[{id:"110471",title:"Dr.",name:"Amjad",middleName:"Zaki",surname:"Almusaed",slug:"amjad-almusaed",fullName:"Amjad Almusaed"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"9376",title:"Contemporary Developments and Perspectives in International Health Security",subtitle:"Volume 1",isOpenForSubmission:!1,hash:"b9a00b84cd04aae458fb1d6c65795601",slug:"contemporary-developments-and-perspectives-in-international-health-security-volume-1",bookSignature:"Stanislaw P. Stawicki, Michael S. Firstenberg, Sagar C. Galwankar, Ricardo Izurieta and Thomas Papadimos",coverURL:"https://cdn.intechopen.com/books/images_new/9376.jpg",editors:[{id:"181694",title:"Dr.",name:"Stanislaw P.",middleName:null,surname:"Stawicki",slug:"stanislaw-p.-stawicki",fullName:"Stanislaw P. Stawicki"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}},{type:"book",id:"7769",title:"Medical Isotopes",subtitle:null,isOpenForSubmission:!1,hash:"f8d3c5a6c9a42398e56b4e82264753f7",slug:"medical-isotopes",bookSignature:"Syed Ali Raza Naqvi and Muhammad Babar Imrani",coverURL:"https://cdn.intechopen.com/books/images_new/7769.jpg",editors:[{id:"259190",title:"Dr.",name:"Syed Ali Raza",middleName:null,surname:"Naqvi",slug:"syed-ali-raza-naqvi",fullName:"Syed Ali Raza Naqvi"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}}],latestBooks:[{type:"book",id:"7434",title:"Molecular Biotechnology",subtitle:null,isOpenForSubmission:!1,hash:"eceede809920e1ec7ecadd4691ede2ec",slug:"molecular-biotechnology",bookSignature:"Sergey Sedykh",coverURL:"https://cdn.intechopen.com/books/images_new/7434.jpg",editedByType:"Edited by",editors:[{id:"178316",title:"Ph.D.",name:"Sergey",middleName:null,surname:"Sedykh",slug:"sergey-sedykh",fullName:"Sergey Sedykh"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"8545",title:"Animal Reproduction in Veterinary Medicine",subtitle:null,isOpenForSubmission:!1,hash:"13aaddf5fdbbc78387e77a7da2388bf6",slug:"animal-reproduction-in-veterinary-medicine",bookSignature:"Faruk Aral, Rita Payan-Carreira and Miguel Quaresma",coverURL:"https://cdn.intechopen.com/books/images_new/8545.jpg",editedByType:"Edited by",editors:[{id:"25600",title:"Prof.",name:"Faruk",middleName:null,surname:"Aral",slug:"faruk-aral",fullName:"Faruk Aral"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"9569",title:"Methods in Molecular Medicine",subtitle:null,isOpenForSubmission:!1,hash:"691d3f3c4ac25a8093414e9b270d2843",slug:"methods-in-molecular-medicine",bookSignature:"Yusuf Tutar",coverURL:"https://cdn.intechopen.com/books/images_new/9569.jpg",editedByType:"Edited by",editors:[{id:"158492",title:"Prof.",name:"Yusuf",middleName:null,surname:"Tutar",slug:"yusuf-tutar",fullName:"Yusuf Tutar"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"9839",title:"Outdoor Recreation",subtitle:"Physiological and Psychological Effects on Health",isOpenForSubmission:!1,hash:"5f5a0d64267e32567daffa5b0c6a6972",slug:"outdoor-recreation-physiological-and-psychological-effects-on-health",bookSignature:"Hilde G. Nielsen",coverURL:"https://cdn.intechopen.com/books/images_new/9839.jpg",editedByType:"Edited by",editors:[{id:"158692",title:"Ph.D.",name:"Hilde G.",middleName:null,surname:"Nielsen",slug:"hilde-g.-nielsen",fullName:"Hilde G. Nielsen"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"7802",title:"Modern Slavery and Human Trafficking",subtitle:null,isOpenForSubmission:!1,hash:"587a0b7fb765f31cc98de33c6c07c2e0",slug:"modern-slavery-and-human-trafficking",bookSignature:"Jane Reeves",coverURL:"https://cdn.intechopen.com/books/images_new/7802.jpg",editedByType:"Edited by",editors:[{id:"211328",title:"Prof.",name:"Jane",middleName:null,surname:"Reeves",slug:"jane-reeves",fullName:"Jane Reeves"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"8063",title:"Food Security in Africa",subtitle:null,isOpenForSubmission:!1,hash:"8cbf3d662b104d19db2efc9d59249efc",slug:"food-security-in-africa",bookSignature:"Barakat Mahmoud",coverURL:"https://cdn.intechopen.com/books/images_new/8063.jpg",editedByType:"Edited by",editors:[{id:"92016",title:"Dr.",name:"Barakat",middleName:null,surname:"Mahmoud",slug:"barakat-mahmoud",fullName:"Barakat Mahmoud"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"10118",title:"Plant Stress Physiology",subtitle:null,isOpenForSubmission:!1,hash:"c68b09d2d2634fc719ae3b9a64a27839",slug:"plant-stress-physiology",bookSignature:"Akbar Hossain",coverURL:"https://cdn.intechopen.com/books/images_new/10118.jpg",editedByType:"Edited by",editors:[{id:"280755",title:"Dr.",name:"Akbar",middleName:null,surname:"Hossain",slug:"akbar-hossain",fullName:"Akbar Hossain"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"9157",title:"Neurodegenerative Diseases",subtitle:"Molecular Mechanisms and Current Therapeutic Approaches",isOpenForSubmission:!1,hash:"bc8be577966ef88735677d7e1e92ed28",slug:"neurodegenerative-diseases-molecular-mechanisms-and-current-therapeutic-approaches",bookSignature:"Nagehan Ersoy Tunalı",coverURL:"https://cdn.intechopen.com/books/images_new/9157.jpg",editedByType:"Edited by",editors:[{id:"82778",title:"Ph.D.",name:"Nagehan",middleName:null,surname:"Ersoy Tunalı",slug:"nagehan-ersoy-tunali",fullName:"Nagehan Ersoy Tunalı"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"9961",title:"Data Mining",subtitle:"Methods, Applications and Systems",isOpenForSubmission:!1,hash:"ed79fb6364f2caf464079f94a0387146",slug:"data-mining-methods-applications-and-systems",bookSignature:"Derya Birant",coverURL:"https://cdn.intechopen.com/books/images_new/9961.jpg",editedByType:"Edited by",editors:[{id:"15609",title:"Dr.",name:"Derya",middleName:null,surname:"Birant",slug:"derya-birant",fullName:"Derya Birant"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}},{type:"book",id:"8686",title:"Direct Torque Control Strategies of Electrical Machines",subtitle:null,isOpenForSubmission:!1,hash:"b6ad22b14db2b8450228545d3d4f6b1a",slug:"direct-torque-control-strategies-of-electrical-machines",bookSignature:"Fatma Ben Salem",coverURL:"https://cdn.intechopen.com/books/images_new/8686.jpg",editedByType:"Edited by",editors:[{id:"295623",title:"Associate Prof.",name:"Fatma",middleName:null,surname:"Ben Salem",slug:"fatma-ben-salem",fullName:"Fatma Ben Salem"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}}]},subject:{topic:{id:"1254",title:"Mobile Robot",slug:"android-science-mobile-robot",parent:{title:"Android Science",slug:"android-science"},numberOfBooks:1,numberOfAuthorsAndEditors:59,numberOfWosCitations:48,numberOfCrossrefCitations:29,numberOfDimensionsCitations:52,videoUrl:null,fallbackUrl:null,description:null},booksByTopicFilter:{topicSlug:"android-science-mobile-robot",sort:"-publishedDate",limit:12,offset:0},booksByTopicCollection:[{type:"book",id:"1881",title:"Mobile Robots",subtitle:"Current Trends",isOpenForSubmission:!1,hash:"6f1ee45d3e50f6d5295a1d8c190b646c",slug:"mobile-robots-current-trends",bookSignature:"Zoran Gacovski",coverURL:"https://cdn.intechopen.com/books/images_new/1881.jpg",editedByType:"Edited by",editors:[{id:"89211",title:"Dr.",name:"Zoran",middleName:null,surname:"Gacovski",slug:"zoran-gacovski",fullName:"Zoran Gacovski"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter",authoredCaption:"Edited by"}}],booksByTopicTotal:1,mostCitedChapters:[{id:"22300",doi:"10.5772/26512",title:"Influence of the Size Factor of a Mobile Robot Moving Toward a Human on Subjective Acceptable Distance",slug:"influence-of-the-size-factor-of-a-mobile-robot-moving-toward-a-human-on-subjective-acceptable-distan",totalDownloads:1626,totalCrossrefCites:4,totalDimensionsCites:10,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Yutaka Hiroi and Akinori Ito",authors:[{id:"66916",title:"Dr.",name:"Yutaka",middleName:null,surname:"Hiroi",slug:"yutaka-hiroi",fullName:"Yutaka Hiroi"},{id:"71987",title:"Prof.",name:"Akinori",middleName:null,surname:"Ito",slug:"akinori-ito",fullName:"Akinori Ito"}]},{id:"22305",doi:"10.5772/25497",title:"Dynamic Modeling and Power Modeling of Robotic Skid-Steered Wheeled Vehicles",slug:"dynamic-modeling-and-power-modeling-of-robotic-skid-steered-wheeled-vehicles",totalDownloads:3951,totalCrossrefCites:6,totalDimensionsCites:7,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Wei Yu, Emmanuel Collins and Oscar Chuy",authors:[{id:"29474",title:"Prof.",name:"Emmanuel",middleName:null,surname:"Collins",slug:"emmanuel-collins",fullName:"Emmanuel Collins"},{id:"39957",title:"Dr.",name:"Oscar",middleName:null,surname:"Chuy",slug:"oscar-chuy",fullName:"Oscar Chuy"},{id:"63522",title:"Dr.",name:"Wei",middleName:null,surname:"Yu",slug:"wei-yu",fullName:"Wei Yu"}]},{id:"22308",doi:"10.5772/25936",title:"Design and Prototyping of Autonomous Ball Wheel Mobile Robots",slug:"design-and-prototyping-of-autonomous-ball-wheel-mobile-robots",totalDownloads:2415,totalCrossrefCites:6,totalDimensionsCites:7,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"H. Ghariblu, A. Moharrami and B. Ghalamchi",authors:[{id:"65019",title:"Dr.",name:"Hashem",middleName:null,surname:"Ghariblu",slug:"hashem-ghariblu",fullName:"Hashem Ghariblu"},{id:"72044",title:"Ms.",name:"Ali",middleName:null,surname:"Moharrami",slug:"ali-moharrami",fullName:"Ali Moharrami"},{id:"72048",title:"Ms.",name:"Behnam",middleName:null,surname:"Ghalamchi",slug:"behnam-ghalamchi",fullName:"Behnam Ghalamchi"}]}],mostDownloadedChaptersLast30Days:[{id:"22298",title:"Mobile Platform with Leg-Wheel Mechanism for Practical Use",slug:"mobile-platform-with-leg-wheel-mechanism-for-practical-use",totalDownloads:3178,totalCrossrefCites:1,totalDimensionsCites:1,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Shuro Nakajima",authors:[{id:"63562",title:"Prof.",name:"Shuro",middleName:null,surname:"Nakajima",slug:"shuro-nakajima",fullName:"Shuro Nakajima"}]},{id:"22292",title:"Autonomous Mobile Robot Emmy III",slug:"autonomous-mobile-robot-emmy-iii",totalDownloads:2069,totalCrossrefCites:0,totalDimensionsCites:1,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Cláudio Rodrigo Torres, Jair Minoro Abe, Germano Lambert-Torres and João Inácio da Silva Filho",authors:[{id:"65919",title:"Dr.",name:"Claudio",middleName:"Rodrigo",surname:"Torres",slug:"claudio-torres",fullName:"Claudio Torres"},{id:"112971",title:"Prof.",name:"Germano",middleName:null,surname:"Lambert-Torres",slug:"germano-lambert-torres",fullName:"Germano Lambert-Torres"},{id:"137199",title:"Dr.",name:"Jair Minoro",middleName:null,surname:"Abe",slug:"jair-minoro-abe",fullName:"Jair Minoro Abe"},{id:"137200",title:"Dr.",name:"João Inácio",middleName:null,surname:"da Silva Filho",slug:"joao-inacio-da-silva-filho",fullName:"João Inácio da Silva Filho"}]},{id:"22293",title:"Mobile Robotics in Education and Research",slug:"mobile-robotics-in-education-and-research",totalDownloads:5113,totalCrossrefCites:0,totalDimensionsCites:5,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Georgios A. Demetriou",authors:[{id:"66153",title:"Dr.",name:"Georgios A.",middleName:null,surname:"Demetriou",slug:"georgios-a.-demetriou",fullName:"Georgios A. Demetriou"}]},{id:"22300",title:"Influence of the Size Factor of a Mobile Robot Moving Toward a Human on Subjective Acceptable Distance",slug:"influence-of-the-size-factor-of-a-mobile-robot-moving-toward-a-human-on-subjective-acceptable-distan",totalDownloads:1626,totalCrossrefCites:4,totalDimensionsCites:10,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Yutaka Hiroi and Akinori Ito",authors:[{id:"66916",title:"Dr.",name:"Yutaka",middleName:null,surname:"Hiroi",slug:"yutaka-hiroi",fullName:"Yutaka Hiroi"},{id:"71987",title:"Prof.",name:"Akinori",middleName:null,surname:"Ito",slug:"akinori-ito",fullName:"Akinori Ito"}]},{id:"22294",title:"The KCLBOT: A Framework of the Nonholonomic Mobile Robot Platform Using Double Compass Self-Localisation",slug:"the-kclbot-a-framework-of-the-nonholonomic-mobile-robot-platform-using-double-compass-self-localisat",totalDownloads:1861,totalCrossrefCites:0,totalDimensionsCites:0,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Evangelos Georgiou, Jian Dai and Michael Luck",authors:[{id:"63454",title:"MSc.",name:"Evangelos",middleName:null,surname:"Georgiou",slug:"evangelos-georgiou",fullName:"Evangelos Georgiou"},{id:"68156",title:"Prof.",name:"Jian",middleName:null,surname:"Dai",slug:"jian-dai",fullName:"Jian Dai"},{id:"68157",title:"Prof.",name:"Michael",middleName:null,surname:"Luck",slug:"michael-luck",fullName:"Michael Luck"}]},{id:"22307",title:"The Development of the Omnidirectional Mobile Home Care Robot",slug:"the-development-of-the-omnidirectional-mobile-home-care-robot",totalDownloads:1994,totalCrossrefCites:0,totalDimensionsCites:1,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Jie-Tong Zou",authors:[{id:"67765",title:"Prof.",name:"Jie-Tong",middleName:"Tong",surname:"Zou",slug:"jie-tong-zou",fullName:"Jie-Tong Zou"}]},{id:"22299",title:"A Micro Mobile Robot with Suction Cups in the Abdominal Cavity for NOTES",slug:"a-micro-mobile-robot-with-suction-cups-in-the-abdominal-cavity-for-notes",totalDownloads:1890,totalCrossrefCites:0,totalDimensionsCites:0,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Chika Hiroki and Wenwei Yu",authors:[{id:"48648",title:"Prof.",name:"Wenwei",middleName:null,surname:"Yu",slug:"wenwei-yu",fullName:"Wenwei Yu"},{id:"71018",title:"Ms.",name:"Chika",middleName:null,surname:"Hiroki",slug:"chika-hiroki",fullName:"Chika Hiroki"}]},{id:"22301",title:"Development of Mobile Robot Based on I2C Bus System",slug:"development-of-mobile-robot-based-on-i2c-bus-system",totalDownloads:3123,totalCrossrefCites:0,totalDimensionsCites:0,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Surachai Panich",authors:[{id:"5704",title:"Dr.",name:"Surachai",middleName:null,surname:"Panich",slug:"surachai-panich",fullName:"Surachai Panich"}]},{id:"22305",title:"Dynamic Modeling and Power Modeling of Robotic Skid-Steered Wheeled Vehicles",slug:"dynamic-modeling-and-power-modeling-of-robotic-skid-steered-wheeled-vehicles",totalDownloads:3951,totalCrossrefCites:6,totalDimensionsCites:7,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Wei Yu, Emmanuel Collins and Oscar Chuy",authors:[{id:"29474",title:"Prof.",name:"Emmanuel",middleName:null,surname:"Collins",slug:"emmanuel-collins",fullName:"Emmanuel Collins"},{id:"39957",title:"Dr.",name:"Oscar",middleName:null,surname:"Chuy",slug:"oscar-chuy",fullName:"Oscar Chuy"},{id:"63522",title:"Dr.",name:"Wei",middleName:null,surname:"Yu",slug:"wei-yu",fullName:"Wei Yu"}]},{id:"22295",title:"Gaining Control Knowledge Through an Applied Mobile Robotics Course",slug:"gaining-control-knowledge-through-an-applied-mobile-robotics-course",totalDownloads:1938,totalCrossrefCites:1,totalDimensionsCites:1,book:{slug:"mobile-robots-current-trends",title:"Mobile Robots",fullTitle:"Mobile Robots - Current Trends"},signatures:"Lluís Pacheco, Ningsu Luo, Inès Ferrer, Xavier Cufí and Roger Arbuse",authors:[{id:"30281",title:"Dr.",name:"Lluís",middleName:null,surname:"Pacheco",slug:"lluis-pacheco",fullName:"Lluís Pacheco"},{id:"32046",title:"Dr.",name:"Ningsu",middleName:null,surname:"Luo",slug:"ningsu-luo",fullName:"Ningsu Luo"},{id:"118672",title:"Dr.",name:"Xavier",middleName:null,surname:"Cufí",slug:"xavier-cufi",fullName:"Xavier Cufí"},{id:"137380",title:"Dr.",name:"Inès",middleName:null,surname:"Ferrer",slug:"ines-ferrer",fullName:"Inès Ferrer"},{id:"137381",title:"Mr.",name:"Roger",middleName:null,surname:"Arbuse",slug:"roger-arbuse",fullName:"Roger Arbuse"}]}],onlineFirstChaptersFilter:{topicSlug:"android-science-mobile-robot",limit:3,offset:0},onlineFirstChaptersCollection:[],onlineFirstChaptersTotal:0},preDownload:{success:null,errors:{}},aboutIntechopen:{},privacyPolicy:{},peerReviewing:{},howOpenAccessPublishingWithIntechopenWorks:{},sponsorshipBooks:{sponsorshipBooks:[{type:"book",id:"10176",title:"Microgrids and Local Energy Systems",subtitle:null,isOpenForSubmission:!0,hash:"c32b4a5351a88f263074b0d0ca813a9c",slug:null,bookSignature:"Prof. Nick Jenkins",coverURL:"https://cdn.intechopen.com/books/images_new/10176.jpg",editedByType:null,editors:[{id:"55219",title:"Prof.",name:"Nick",middleName:null,surname:"Jenkins",slug:"nick-jenkins",fullName:"Nick Jenkins"}],equalEditorOne:null,equalEditorTwo:null,equalEditorThree:null,productType:{id:"1",chapterContentType:"chapter"}}],offset:8,limit:8,total:1},route:{name:"profile.detail",path:"/profiles/204097/bakiye-goker-bagca",hash:"",query:{},params:{id:"204097",slug:"bakiye-goker-bagca"},fullPath:"/profiles/204097/bakiye-goker-bagca",meta:{},from:{name:null,path:"/",hash:"",query:{},params:{},fullPath:"/",meta:{}}}},function(){var e;(e=document.currentScript||document.scripts[document.scripts.length-1]).parentNode.removeChild(e)}()