Image Analysis Of Microstructure Research Articles

Deep Learning Methods for Microstructural Image Analysis: The State-of-the-Art and Future Perspectives

Deep Learning Methods for Microstructural Image Analysis: The State-of-the-Art and Future Perspectives

Biochemical, Microstructural, and Probiotic Bacterial Patterns of Innovative Fresh Cheese Fortified with Helianthus tuberosus Tubers

Recently, functional foods have become the aim of customers and food producers. Integrating vegetable ingredients in the food industry is a productive goal to reduce costs while maintaining quality. Dried Jerusalem artichoke tubers powder (DJATP) was used as a novel approach in cheese manufacturing. Innovatively, its holding capacity features and impact on probiotic development were evaluated. The SEM microstructure image and biochemical analysis of DJATP confirmed higher water holding (5.31 g/g), oil absorption (1.99 g/g), and swelling (1.79 g/g) capacities than casein. DJATP (3%) supported the probiotic bacterial growth (Streptococcus thermophilus, Bifidobacterium bifidum, and Lactobacillus acidophilus) and accelerated the fermentation of skimmed milk more than pure inulin. When fortified with DJATP (3% or 6%), the cheese yield increased (24.66% and 27.85%, respectively) compared with 17.55% for control after storage (14 days). Besides the high levels of amino acids, minerals, flavonoids, phenols, and antioxidants, the probiotic bacteria in the DJATP-fortified cheese were better active, with better sensory features, recording the highest judging score (87.67) against the control (79.00). To our knowledge, no preceding studies used DJATP in fresh cheese manufacturing followed the probiotic behavior in DJATP or compared the microstructure of DJATP and casein. Instead of inulin, our novel approach suggests using DJATP as a prebiotic and an enhancer for fresh cheese quality and yield, all while being cost-effective. Future studies are encouraged to explore the potential use of DJATP in other functional cheese products.

Machine Learning-Based Prediction of Grain Size from Colored Microstructure

We constructed a convolutional neural network to estimate average grain size from microstructure images. In the previous study from our research group, the network was trained using GB-type images in which the grain matrix and grain boundary were represented in white and black, respectively. The model well estimated the same GB-type images, but did not properly predict CL-type images where grain boundaries were defined by color contrast between grains. In the present study, the convolutional neural network was trained using CL-type microstructure images, and evaluated the average grain size, for comparison with the previous GB-type model. The relationship between the microstructure image and the average grain size was determined using regression rather than classification. Then, the results were compared with the previous ones. Finally, the proposed approach was used for actual microstructural image analysis. Mid-layer images were extracted to examine how the network recognizes the characteristics of microstructures, such as grain color and grain boundary. Like the previous GB-type model, the present CL-type model seems to estimate the average grain size from the curvature of the grain boundary through edge detection of the grain boundaries. However, the GB- and CL-type models only properly predicted the grain size from the same kind of images as the training data, because the definitions of the grain boundaries of the two models were different.

Innovative sieve simulation and microstructure image analysis techniques for estimation of aggregate size distribution in hardened concrete

Aggregate size distribution is one of the most important parameters affecting the strength and workability of concrete. Therefore, estimating aggregate size distribution from hardened concrete samples can be utilized to evaluate the quality of the concrete mixture. For this purpose, an innovative and comprehensive method was developed in which images of the macrostructure and microstructure of hardened concrete were utilized. Macrostructure images were analyzed to extract coarser aggregates (retained on the 0.3 mm sieve). A simulation of the mechanical sieving was developed to assess the passing of each aggregate based on relative translational and rotational motions of the aggregate image with respect to the apertures. This simulation can be claimed as the most accurate method for examining the passing of aggregates through sieves using 2D images. In addition, some geometric rules were hired to optimize the computational costs of the simulation. In order to determine the volume ratio of finer aggregates (passing the 0.3 mm sieve), two relationships between the mass and volume of concrete phases and hardened cement components were considered. Finally, the volume ratio of finer aggregates was estimated by using these relationships and by analyzing microstructure images containing high resolution image for calculating air voids area ratio and BSE image for calculating hardened cement components area ratios. Comparison of the results from the proposed method with actual grading highlights the brilliant performance of the proposed method in estimating aggregate size distribution.

Automatic Segmentation of Osteonal Microstructure in Human Cortical Bone Using Deep Learning: A Proof of Concept.

Cortical bone microstructure assessment in biological and forensic anthropology can assist with the estimation of age-at-death and animal-human differentiation, for example. Osteonal structures within cortical bone are the key feature under analysis, with osteon frequency and metric parameters providing crucial information for the assessment. Currently, the histomorphological assessment consists of a time-consuming manual process for which specific training is required. Our work investigates the feasibility of automatic analysis of human bone microstructure images through the application of deep learning. In this paper, we use a U-Net architecture to address the semantic segmentation of such images into three classes: intact osteons, fragmentary osteons, and background. Data augmentation was used to avoid overfitting. We evaluated our fully automatic approach using a sample of 99 microphotographs. The contours of intact and fragmentary osteons were traced manually to provide ground truth. The Dice coefficients were 0.73 for intact osteons, 0.38 for fragmented osteons, and 0.81 for background, giving an average of 0.64. The Dice coefficient of the binary classification osteon-background was 0.82. Although further refinement of the initial model and tests with larger datasets are needed, this study provides, to the best of our knowledge, the first proof of concept for the use of computer vision and deep learning for differentiating both intact and fragmentary osteons in human cortical bone. This approach has the potential to widen and facilitate the use of histomorphological assessment in the biological and forensic anthropology communities.

Porous beta titanium alloy coated with a therapeutic biopolymeric composite to improve tribomechanical and biofunctional balance

Tooth loss is common in patients struggling with dental cavities, periodontal diseases, and tumors, as well as those who abuse tobacco or drugs. In this scenario, dental implants have become the primary treatment option for complete or partial tooth loss. Dental implant failure can be caused by stress shielding phenomenon, poor osseointegration, or to bacterial infections. In the present study, a joint solution to these limitations is proposed using a variety of porous β-titanium substrates using powder Ti35Nb7Zr5Ta alloy and employing the spacer-holder approach (ammonium bicarbonate) to obtain a variety of porosity percentage (30, 40, and 50 vol%), and pore diameters in 100–200 μm, that has been characterized in terms of its size distribution, density, morphology, chemical composition, compaction ability and Vickers micro-hardness. Furthermore, porosity, microstructure (Archimedes and image analysis) and tribomechanical behavior (P-h curves and scratch tests) experiments were performed to study and characterize the porous substrates. Polyvinyl alcohol (PVA)/poly-ε-caprolactone (PCL) containing silver nanoparticles (AgNPs), as antibacterial composite, was employed to infiltrate β-Ti disks. Scanning electron microscopy was used to determine the coating morphology, thickness, and infiltration of the porous substrates. Wettability and SBF experiments were also carried out to investigate hydrophobicity and potential bio-functionality. The results suggested how the porosity of the β-Ti alloy affects the mechanical characteristics and the wettability of the substrate that was successfully infiltrated to exert an antimicrobial behavior.

Modelling Effects of Copper on Microstructure and Properties of Al–Si–Cu Alloys

The combined impacts of copper content, cold working and homogenization in aluminium–silicon alloys were examined using microstructure study, thermodynamic modelling, image analysis and hardness measurement. Two aluminium alloys of varying copper contents were cast in a permanent mould using induction furnace. Microstructural observation revealed the formation of varying amount of intermetallic phases due to the presence of copper. Image analysis confirms the relative changes in percent phase fractions owing to presence of copper as intermetallic phases increases. Al2Cu phase was most favoured with increasing copper. Cold working was performed by reduction of 5%, 10% and 15% of the cast alloys by equal amounts of strain. Hardness measurements indicated increase in hardness values attributed to both higher amounts of copper and cold working. Al2Cu phase was predicted contribute compared to Al7Cu4Ni and Alpha phases. Homogenization at 300°C for 4 hr would suffice to obtain a uniform structure. The reasons for changes in microstructure and mechanical properties for different copper content were established by thermodynamic phase fraction calculation. Higher amounts of copper leads to greater volume fractions of intermetallic phases. Using solidification calculations, it is found that copper in liquid encourages dendrite formation which was also evident from microstructure.

Influence of Equal Channel Angular Extrusion on the Behavior of Lead Alloy

Equal Channel Angular Pressing (ECAP) is the most promising material processing technique involving severe plastic deformation, and has been extensively employed and analysed. The aim of this work is to examine the influence of ECAP on the behaviour of Lead alloy. The technique was applied to Lead alloy at room temperature using route C, at channel angles of 450, 600, 750, 900 and 1050. The materials were processed up to five ECAP passes. Hardness test, impact test, and microstructural changes of the processed materials were examined. Results show that extrusion force reduces as the strain level increases and that dynamic recrystallization and structural changes reduce the material hardness for all angles of ECAP. All the angles absorbed their least amount of energy at their 5th pass. Analysis of microstructure images also revealed that increasing the strain level leads to break down and dissolution of Antimony rich precipitate.

Quantitative image analysis of protein foam microstructure and its correlation with rheological properties: Egg white foam

Recent advances in software development and machine learning algorithms are revolutionizing the way microstructures are analyzed and quantified in fields like biology and neuroscience. With this, comes an upsurge in the opportunities to apply these tools to study food microstructure to gain a deeper understanding of food structure-function relationships. This article shows how a recently developed deep-learning cellular segmentation algorithm, ‘ Cellpose ’ can be used to identify foam microstructure for further quantification. It successfully identified the air bubbles in a protein foam matrix, from microscopic images captured on a simple brightfield microscope. The segmentation algorithm allowed further quantification of microstructural parameters of the air phase (bubbles) and of the liquid phase (lamella) of the foams. Egg white foams were made with basic ingredients for meringue and the effect of sugar concentration and acidic conditions were studied. Microstructural parameters were analyzed in relation to the rheological responses of the foams. The underlying microstructural mechanisms governing the changes in the foams' stiffness and linearity of the viscoelastic response are shown. Sugar changes bubble size distribution by thickening liquid egg white, it also shortened the linear response to deformation due to its decreasing lamella thickness. Acidity re-shaped the bubbles into more ‘hexagon-like’ structures allowing more air to be incorporated in the foams. The shape of the bubbles under acidic conditions also makes the foams extend their linear response to deformation. The understanding achieved with data from the algorithm-identified microstructures presents a new way to characterize and study food foams. • Microscopy images of egg white foam were captured using a simple brightfield microscope. • The air and liquid phases were successfully distinguished by the Cellpose algorithm. • Quantitative microstructural parameters of both phases were obtained in subsequent image analyses. • Sugar decreases bubble size while acidity re-shapes bubbles into a hexagon-like shape. • Sugar and acid microstructural mechanisms alter the linear viscoelastic response of the foams. • Obtaining microstructural paraments leads to new insights into the study of foams.

Estimation of Average Grain Size from Microstructure Image Using a Convolutional Neural Network.

In this study, the average grain size was evaluated from a microstructure image using a convolutional neural network. Since the grain size in a microstructure image can be directly measured and verified in the original image, unlike the chemical composition or mechanical properties of material, it is more appropriate to validate the training results quantitatively. An analysis of microstructure images, such as grain size, can be performed manually or using image analysis software; however, it is expected that the analysis would be simpler and faster with machine learning. Microstructure images were created using a phase-field simulation, and machine learning was carried out with a convolutional neural network model. The relationship between the microstructure image and the average grain size was not judged by classification, as the goal was to have different results for each microstructure using regression. The results showed high accuracy within the training range. The average grain sizes of experimental images with explicit grain boundary were well estimated by the network. The mid-layer image was analyzed to examine how the network understood the input microstructure image. The network seemed to recognize the curvatures of the grain boundaries and estimate the average grain size from these curvatures.

About the experience of using hardmetals in the production of roller cone bits at Volgaburmash JSC

Since 2017, Volgaburmash JSC (Samara, Russia) tested purchased 90%WC–10%Co carbide powder mixtures and finished carbide drill bits from various manufacturers. The work was carried out in order to check the possibility of using purchased products as raw materials at the plant to reduce the production cycle for the manufacture of carbide inserts for roller cone bits. This intercommodity substitution (outsourcing) is carried out with the aim of potential cone bit cost reduction and production process acceleration so that the plant can operate in the heavily competitive environment of foreign and domestic markets. The article focuses on the analysis and detailed comparison of the micro- and macrostructure, physical, mechanical, chemical and processing properties of purchased hard-alloy mixtures and sintered inserts of various manufacturers including Volgaburmash JSC. All properties of materials under study were determined in accordance with the VBM JSC company standard STP 582-17. Much attention is paid to comparing crack resistance or Palmqvist fracture toughness values of the alloy and analysis of microstructure images and fracture propagation pattern after using scanning electron microscopy tests. In addition, consideration is given to such important hard alloy properties as hardness and transverse bending strength. Based on the results of the conducted research, conclusions are presented on the expediency of using purchased hard-alloy materials at the Volgaburmash JSC metallurgical shop in comparison with internally manufactured materials.

Numerical modeling for particulate flow through realistic microporous structure of microfiltration membrane: Direct numerical simulation coordinated with focused ion beam scanning electron microscopy

In this study, the focused ion beam scanning electron microscope (FIB-SEM) technique is utilized to obtain a realistic microstructure geometry of mullite and α-alumina membranes, which are commercially available microfiltration (MF) membranes, for implementation in a high-resolution numerical simulation model based on coupled direct numerical simulation (DNS) and the discrete element method (DEM). The flow and particle permeation through the membrane pores are evaluated for different particle sizes. The developed simulation model is a powerful tool for enhancing our understanding of the MF process because it introduces the realistic microporous structure of an actual MF membrane rather than the simplified membrane model structures such as straight cylindrical pores and particle packing structures that have been used in previous numerical studies. The application of the developed model is discussed for the prediction of the rejection curve, which shows the importance of considering a realistic microporous structure for the accurate prediction of the MF process. The results showed different rejection curves for the mullite and α-alumina membranes, which mainly contributed to the difference in the local pore size distributions of the two membranes. Furthermore, detailed particle permeation through the membrane is evaluated, which is not feasible in experimental analyses for realistic MF membranes. • Realistic microfiltration geometry was obtained using FIB-SEM technique. • High resolution DNS-DEM was developed for detailed particle permeation analysis. • Different rejection curve for mullite and α-alumina membranes was obtained. • Membrane microstructure affected the particle permeation behavior and rejection. • Image analysis of microstructure supported the DNS-DEM results.

Improvements on hot corrosion behaviour of HVOF coated CO2 laser beam and pulsed gas tungsten arc weldments in 2.5% sulphur gas plus molten salt in industrial waste incinerator environment

This research article discusses in detail how the‌ ‌rate‌ ‌of‌ ‌corrosion‌ ‌increases‌ for the Alloy 254 in the ‌molten‌ ‌salt‌ ‌mixture‌ ‌of‌ ‌10‌ ‌wt%‌ ‌NaCl‌ ‌+ 40‌ ‌wt%‌ ‌K‌2‌SO‌4‌ + 40‌ ‌wt%‌ ‌Na‌2‌SO‌4‌ ‌+‌ ‌10‌ ‌wt%‌ ‌KCl‌ ‌and‌ ‌corrosive‌ ‌gas‌ combination of ‌ ‌2.5‌%‌ ‌SO‌2‌ ‌+‌ ‌3‌%‌ ‌O‌2‌ ‌+‌ ‌25‌%‌ ‌CO‌2‌ ‌+‌ ‌69.5‌%‌ ‌N‌2 in simulated municipal‌ ‌solid‌ ‌waste‌ ‌‌incinerator environment at 700 °C and 800 °C.‌ ‌In addition, the article shows rate of ‌thermal corrosion abridged due to the addition‌ ‌of‌ ‌80‌ ‌Ni-‌ ‌20‌ ‌Cr‌ ‌-‌High-velocity oxygen fuel ‌coating‌ ‌on‌ ‌the‌ ‌SMO‌ ‌254‌ ‌specimen. Thermogravimetric analysis is adapted to analyze the corrosion rate. A‌ ‌higher‌ ‌hardness‌ ‌was‌ ‌obtained‌ ‌for‌ ‌the‌ ‌High-velocity oxygen fuel coated‌ ‌specimen‌ ‌than‌ ‌that‌ ‌of‌ ‌base‌ ‌metal.‌ ‌Surface‌ ‌roughness‌ ‌was‌ ‌measured‌ ‌on‌ ‌the‌ ‌thermal‌ ‌barrier‌ ‌coated‌ ‌surface.‌ As a result of the impact of molten salt and corrosive gas, hot corrosion by-products such as ‌sulphides, chlorides‌ ‌and‌ ‌Na‌ are produced, which are observed to increase the severity of corrosion rate. ‌‌ The study provides a detailed analysis of the metallurgical changes along the cross-section on the hot corroded samples by optical and Scanning electron microstructural image analysis. X-ray diffraction analysis was used to know about the corrosion products formed during corrosion. Formation‌ of the major protective oxide phases such as NiCr2O4, Cr2O3, NiO and CrMnO4 provide better corrosion resistance to the substrate in the coated samples. The diffusion of the corrosive elements along the cross-section of the weld interface, weld zone, and base metal zone was analyzed in detail using Electron dispersive spectroscopy‌ ‌data in the point scan, line scan, and x-ray mapping techniques. Thermal spray coatings were found to be effective to prevent the hot corrosion, however, a higher weight gain (17.75 mg cm−2) was observed on the pulse current gas tungsten arc weldment sample in the simulated environment compared to a weight gain (2.66 mg cm−2) base metal specimen at 700 °C. Spallation of coated surface occurred at 800 °C, and it could be concluded that the coating has found to be effective in preventing the hot corrosion and safe to operate at a temperature of 700 °C for Alloy 254 in this aggressive simulated environment.

Dynamic Characteristics and Mechanism of the Saturated Compacted Loess under Freeze-Thaw Cycles

To study the dynamic characteristics and mechanism of saturated loess after freeze-thaw cycles, a series of laboratory tests including freeze-thaw cycle tests, dynamic triaxial tests, and scanning electron microscope tests of the saturated remolded loess was conducted. The characteristics of the dynamic parameters of the saturated loess after different freeze-thaw cycles were discussed. The characteristics of the microstructure parameters changes were analyzed. The evolution process and mechanism of the microstructure of the remolded loess under freeze-thaw cycles were proposed. The results show that after different freeze-thaw cycles, the dynamic stress-dynamic strain curves of the saturated remolded loess conform to the hyperbolic model; however, the freeze-thaw cycle has a significant effect on the model parameter b . With the increase of freeze-thaw cycles, the dynamic shear modulus of saturated remolded loess first decreases and then increases, while the damping ratio is opposite. When saturated remolded loess experiences freeze-thaw cycles greater than four, its dynamic stability is better than that of saturated soil without freeze-thaw cycles. The dynamic stability reaches its peak after seven freeze-thaw cycles and is equivalent to that of saturated soil without freeze-thaw cycles after forty cycles. Combined with the results of the quantitative analysis of microstructure images, with the increase of the freeze-thaw cycles, the number of large and medium particles in the soil reduces, and the number of micros and small particles increases. The particle size tends to be uniform. The apparent porosity increases rapidly and then decreases sharply and tends to be stable after 4 freeze-thaw cycles. The pore and particle fractal dimensions continue to decrease. The probability of entropy increases first and then decreases. It is illustrated that the saturated loess has mainly experienced three steps under freeze-thaw cycles: (1) fracture and expansion of original skeleton cementation, (2) damage, crushing and aggregation of the particle, and (3) compaction and reorganization of soil structure. Besides, the saturation condition significantly accelerates the evolution process of the internal structure of the soil under freeze-thaw cycles. These lead to the strengthening effect of soil dynamic stiffness under long-term freeze-thaw cycles.

Effect of processing speed on wear and corrosion behavior of novel MoS2 and CeO2 reinforced hybrid aluminum matrix composites fabricated by friction stir processing

The extensive use of aluminum matrix composites as structural material urges it to possess good surface properties such as wear and corrosion resistance. To achieve that, mixture of MoS 2 and CeO 2 powder was reinforced in Al7075 alloy through friction stir processing with different processing speed and multi-pass approach. The microstructural image analysis by optical microscope and scanning electron microscope revealed reduction of matrix grain size and improvement of particle distribution for lower processing speed and multi-pass processing. The elemental analysis indicated well dispersion of reinforcement particles in matrix along with the presence of compounds of aluminum alloy. The microhardness and tensile strength were increased with reduction of processing speed and increase of number of pass. The friction coefficient and the wear loss were also low for composites produced with low processing speed and multi-pass. A combined abrasive and adhesive wear was observed in all the processed composites. With increase of processing speed and number of pass, the corrosion behavior of the composites in 3.5% NaCl solution deteriorated. The composite produced with 30 mm/min and 2 pass exhibited best particle distribution, microhardness and wear resistance. However, the composite fabricated with 30 mm/min and 1 pass yielded best corrosion resistance. • Novel hybrid aluminum matrix composites developed by friction stir processing • Microstructure of the fabricated composites analyzed • Microhardness distribution measured • Wear mechanism and friction coefficient variation analyzed and effect of processing speed on wear loss discussed • Corrosion behavior and effect of process parameters on corrosion parameters discussed

Characterization of Microstructure of Crept Samples of Dissimilar Weld Joint Using Standard and Advanced Electron Microscopy Techniques

Conventional long-term creep test (CCT) to the rupture and so called accelerated creep test (ACT) of the dissimilar weld joint made of FB2 and F martensitic steels and of the base materials were carried out at temperatures ranging from 550 °C to 650 °C in the stress range from 70 to 220 MPa. Assessment of microstructure development and changes of hardness was correlated with the creep strength. During creep at temperatures above 575 °C Laves phase precipitated in all parts of the weld joint and especially in the heat affected zones. Coarse Laves phase particles and their clusters with chromium carbides served as nucleation centers for cavities. As the fine grained heat affected zone of F steel was the softest part of the weld joint, many cavities originated and cause failure of samples. The aim of this paper is to compare results and possibilities of the “standard” methods and advanced scanning electron microscopy performed by instrument equipped with a concentric backscatter electron detector (CBS). Filtering of the signal enables improving and/or diminishing of selected type of contrast caused by various types of particles of secondary phases. The images were used as an input data for image analysis and developments of microstructures during CCT and ACT were compared. Results have shown that specimens after ACT contains significantly lower content of the Laves phase.

Multi-Parametric Optimisation by Quantitative Assessment of Distribution Index and Area Fraction of Composite

Abstract The mechanical properties of the particulate reinforced composites are profoundly affected by two significant factors, fraction of reinforcement in the matrix and the distribution of the reinforced particles in the matrix. In this paper, a novel approach is adopted where a newly developed distribution index and area fraction of the composites are quantitatively determined through microstructural image analysis method, and the process parameters are optimised for maximum area fraction and minimum distribution index. The distribution index (D index) is calculated by estimating the free-path spacing within particles. The stir casting process parameters (stirrer rpm, time of stirring, furnace temperature and preheat temperatures) are optimised by Box-Behnken design of response surface methodology. The quantitatively assessed responses provide an authentic base for optimisation of process parameters. The experimental responses corresponding to the optimised parameters are within the range of the responses estimated by the confirmation test of optimisation with a 95 % confidence level.

Unravelling Antarctica’s past through the stratigraphy of a deep ice core: an image-analysis study of the EPICA-DML line-scan images

Polar ice research has undergone great progress in the last six decades. One of its recent technological achievements has been the development of new techniques for digital image recording and analysis of ice-core stratigraphy and microstructure. In this work we investigate one such image records, namely the line-scan image records of the EPICA-DML (European Project for Ice Coring in Antarctica, Dronning Maud Land) deep ice core. These images provide a multiscale depiction of the stratigraphy and structure of the Antarctic Ice Sheet. While previous studies have focused on the ice-core optical stratigraphy on the micro- and mesoscale (<1 mm and 10−3–1 m, respectively), in this work we present several methods to obtain fast and reliable information on the ice-core stratigraphy on the macroscale (1–103 m), including the full ice-sheet thickness. The paleoclimatic relevance of the ice-core optical stratigraphy on the macroscale is demonstrated through the comparison of the line-scan grey-value record of the EPICA-DML deep ice core with its mineral dust record, which is used as a proxy for microinclusions and for several other types of climate proxies. Additionally, we introduce a novel method to estimate the macroscopic air-bubble concentration (including number and size of bubbles) in ice cores, which is simpler, faster, and almost as reliable as painstaking microscopic studies. After a brief excursion on the relation between macroscopic and mesoscopic measures of optical stratigraphy, we close this work by making the case for a multi-measure analysis of ice-core line-scan images, which enables us to obtain a broad perspective of the optical stratigraphy of the whole ice core, with relevance for paleoclimate and ice-sheet-flow studies.

Quality parameters and morphometric characterization of hot-air popcorn as related to moisture content

Upon popping, corn kernels undergo substantial dehydration. Initial moisture content of the kernel, popping method and temperature determine the final quality, and physical characteristics of the flakes. In this work, corn kernels (Zea Mays var Everta) at different moisture contents were subjected to hot-air popping and quality, physical and morphometric parameters of the flakes were evaluated, and related with their microstructural features. The best moisture content for hot-air popping was 9.14%. Macro and microstructural examination and digital image analysis of the flakes showed that, at the macro level, the roughness of the different flakes, as given by their fractal dimension (FD), were similar among morphologies, whereas at the microscale, it was possible to differentiate between them. Additionally, at the macro-scale, for the different morphologies high correlations were found between areas and the linear dimensions (Feret or width)FD of evaluated flakes. Moreover, when using the FD of the microstructures as the exponent, high correlations were also observed, which confirms an analog multiscale behavior between the roughness of the flakes and their microstructures.

Machine Learning Versus Human-Developed Algorithms in Image Analysis of Microstructures

Abstract Automatic image analysis is nowadays a standard method in quality control of metallic materials, especially in grain size, graphite shape and non-metallic content evaluation. Automatically prepared solutions, based on machine learning, constitute an effective and sufficiently precise tool for classification. Human-developed algorithms, on the other hand, require much more experience in preparation, but allow better control of factors affecting the final result. Both attempts were described and compared.