This PDF file contains the front matter associated with SPIE Proceedings Volume 9534, including the Title Page, Copyright information, Table of Contents, Authors, Introduction (if any), and Conference Committee listing.

The search for an efficient on-line monitoring system focused on the real-time analysis of arc welding quality is an active area of research. The topography and the superficial temperature field of the weld pool can provide important information which can be used to regulate the welding parameters for depositing consistent welds. One difficulty relies on accessing this information despite the bright dazzling welding arc. In the present work, Stokes polarimetry and associated shape-from-polarization methods are applied for the analysis of the weld pool through its 810 nm-wavelength infrared emissions. The obtained information can provide a better understanding of the process, such as the usage of the topography to seek Marangoni flows direction, or to have a denser 3D map to improve numerical simulation models.

A new interactive approach for segmentation and classification of cementitious materials using Scanning Electron Microscope images is presented in this paper. It is based on the denoising of the data with the Block Matching 3D (BM3D) algorithm, Binary Partition Tree (BPT) segmentation and Support Vector Machines (SVM) classification. The latter two operations are both performed in an interactive way. The BPT provides a hierarchical representation of the spatial regions of the data and, after an appropriate pruning, it yields a segmentation map which can be improved by the user. SVMs are used to obtain a classification map of the image with which the user can interact to get better results. The interactivity is twofold: it allows the user to get a better segmentation by exploring the BPT structure, and to help the classifier to better discriminate the classes. This is performed by improving the representativity of the training set, adding new pixels from the segmented regions to the training samples. This approach performs similarly or better than methods currently used in an industrial environment. The validation is performed on several cement samples, both qualitatively by visual examination and quantitatively by the comparison of experimental results with theoretical values.

The acoustic tomographic technique is used in the diagnosis process of standing trees. This paper presents a segmentation methodology to separate defective regions in cross-section tomographic images obtained with Arbotom® device. A set of experiments was proposed using two trunk samples obtained from a eucalyptus tree, simulating defects by drilling holes with known geometry, size and position and using different number of sensors. Also, tomographic images from trees presenting real defects were studied, by testing two different species with significant internal decay. Tomographic images and photographs from the trunk cross-section were processed to align the propagation velocity data with a corresponding region, healthy or defective. The segmentation was performed by finding a velocity threshold value to separate the defective region; a logistic regression model was fitted to obtain the value that maximizes a performance criterion, being selected the geometric mean. Accuracy segmentation values increased as the number of sensors augmented; also the position influenced the result, obtaining improved results in the case of centric defects.

Texture discrimination was the second more important task studied after colour perception and characterization. Nevertheless, colour texture assessment and characterization was few studied and no vector processing was proposed to assess this important visual information. In this work we show the construction of a new vector that integrates fully the information of texture and color. This vector is based on Julesz psico-physics conjectures and the Haralick cooccurrence matrix. A colour texture image in the CIEL*a* b* colour space is characterizing in a 3D matrix, from which it is possible to visually some variations in chromaticity. The performance of this vector had evaluated in tasks of classification in front of other developments that mix the texture and colour information. The colour contrast occurrence matrix (C2O) has the best classification rates in three of the four image database evaluated as OUTEX, VISTEX, STEX and ALOT. C2O texture classification was evaluated in front of co-occurrence matrix (GLMC), run-length matrix (RLM) and local binary patterns (LBP) approaches.

The research purpose is to improve aesthetic anomalies detection and evaluation based on what is perceived by human eye and on the 2006 CIE report.¹ It is therefore important to define parameters able to discriminate surfaces, in accordance with the perception of human eye. Our starting point in assessing aesthetic anomalies is geometric description such as defined by ISO standard,² i.e. traduce anomalies description with perception words about texture divergence impact. However, human controllers observe (detect) the aesthetic anomaly by its visual effect and interpreter for its geometric description. The research question is how define generic parameters for discriminating aesthetic anomalies, from enhanced information of visual texture such as recent surface visual rendering approach. We propose to use an approach from visual texture processing that quantify spatial variations of pixel for translating changes in color, material and relief. From a set of images from different angles of light which gives us access to the surface appearance, we propose an approach from visual effect to geometrical specifications as the current standards have identified the aesthetic anomalies.

Geometry and texture image decomposition is an important paradigm in image processing. Following to Yves Meyer works based on Total Variation (VT), the decomposition model has known a renewed interest. In this paper, we propose an algorithm which decomposes color image into geometry and texture component by projecting the image in a bivariate polynomial basis and considering the geometry component as the partial reconstruction and the texture component as the remaining part. The experimental results show the adequacy of using our method as a texture extraction tool. Furthermore, we integrate it into a dynamic texture classification process.

This paper deals with print quality control through a spectral color measurement. The aim is to estimate the spectral reflectance curve of each pixel of a printed sheet for a spectral matching with the reference image. The proposed method consists to perform a spectral characterization of the complete chain which includes the printing system and a digital trichromatic camera. First, the spectral printer model is presented and verified by experiments. Then, the camera spectral sensitivity curves are estimated through the capture of a color chart whose spectral reflectance curves have been previously measured. Finally, the spectral printer model is used to estimate the print spectral reflectance curves from camera responses.

In this paper an automatic classification method of skin lesions from dermoscopic images is proposed. This method is based on color texture analysis based both on color mathematical morphology and Kohonen Self-Organizing Maps (SOM), and it does not need any previous segmentation process. More concretely, mathematical morphology is used to compute a local descriptor for each pixel of the image, while the SOM is used to cluster them and, thus, create the texture descriptor of the global image. Two approaches are proposed, depending on whether the pixel descriptor is computed using classical (i.e. spatially invariant) or adaptive (i.e. spatially variant) mathematical morphology by means of the Color Adaptive Neighborhoods (CANs) framework. Both approaches obtained similar areas under the ROC curve (AUC): 0.854 and 0.859 outperforming the AUC built upon dermatologists' predictions (0.792).

Prostate cancer has been reported as the second most frequently diagnosed men cancers in the world. In the last decades, new imaging techniques based on MRI have been developed in order to improve the diagnosis task of radiologists. In practise, diagnosis can be affected by multiple factors reducing the chance to detect potential lesions. Computer-aided detection and computer-aided diagnosis have been designed to answer to these needs and provide help to radiologists in their daily duties. In this study, we proposed an automatic method to detect prostate cancer from a per voxel manner using 3T multi-parametric Magnetic Resonance Imaging (MRI) and a gradient boosting classifier. The best performances are obtained using all multi-parametric information as well as zonal information. The sensitivity and specificity obtained are 94:7% and 93:0%, respectively and an Area Under Curve (AUC) of 0:968.

Computer aided planning for orthodontic treatment requires knowing occlusion of separately scanned dental casts. A visual guided registration is conducted starting by extracting corresponding features in both photographs and 3D scans. To achieve this, dental neck and occlusion surface are firstly extracted by image segmentation and 3D curvature analysis. Then, an iterative registration process is conducted during which feature positions are refined, guided by previously found anatomic edges. The occlusal edge image detection is improved by an original algorithm which follows Canny’s poorly detected edges using a priori knowledge of tooth shapes. Finally, the influence of feature extraction and position optimization is evaluated in terms of the quality of the induced registration. Best combination of feature detection and optimization leads to a positioning average error of 1.10 mm and 2.03°.

Breast cancer is the second most common cancer and the leading cause of cancer death among women. Medical imaging has become an indispensable tool for its diagnosis and follow up. During the last decade, the medical community has promoted to incorporate Ultra-Sound (US) screening as part of the standard routine. The main reason for using US imaging is its capability to differentiate benign from malignant masses, when compared to other imaging techniques. The increasing usage of US imaging encourages the development of Computer Aided Diagnosis (CAD) systems applied to Breast Ultra-Sound (BUS) images. However accurate delineations of the lesions and structures of the breast are essential for CAD systems in order to extract information needed to perform diagnosis. This article proposes a highly modular and flexible framework for segmenting lesions and tissues present in BUS images. The proposal takes advantage of optimization strategies using super-pixels and high-level descriptors, which are analogous to the visual cues used by radiologists. Qualitative and quantitative results are provided stating a performance within the range of the state-of-the-art.

In this paper, we investigated the possibility of getting wide-field images of corneal endothelium for patients. An optical set-up coupled to a numerical reconstruction based on Structured Illumination Mircoscopy (SIM) has been developed in order to isolate the tiny volume which contains the endothelial mono-layer found at the inner surface of the cornea. At this moment, this imaging system seems compromised for patients and further refinement are investigated for stored humans corneas banks.

In this article, manually segmented corneal endothelial mosaic will be characterized with spatial statistical functions and criteria issued from granulometry and morphometry. A novel approach to simulate spatial tessellations with Gaussian random fields with Gaussian and Bessel covariance functions, watershed and h-maxima is reported. Finally, these random spatial tessellations will be characterized and compared to corneal mosaics.

This paper presents a novel approach for automatic pupil segmentation. The proposed algorithm uses local histogram-based threshold, area and eccentricity that looks for the region that has the highest probability of having the pupil. Proposed algorithm is implemented on FPGA using a non-iterative scheme along with hardware optimized median filter and connected component logic algorithm. The proposed algorithm is tested on two public databases namely: CASIA v1.0 and MMU v1.0. Experimental results show that the proposed method has satisfying performance and good robustness against the reflection in the pupil.

Carbon blacks are widely used as filler in industrial products to modify their mechanical, electrical and optical properties. For rubber products, they are the subject of a standard classification system relative to their surface area, particle size and structure. The electron microscope remains the most accurate means of measuring these characteristics on condition that boundaries of aggregates and particles are correctly detected. In this paper, we propose an image processing chain allowing subsequent characterization for automatic grading of the carbon black aggregates. Based on literature review, 31 features are extracted from TEM images to obtain reliable information on the particle size, the shape and microstructure of the carbon black aggregates. Then, they are used for training several classifiers to compare their results for automatic grading. To obtain better results, we suggest to use a cluster identification of aggregates in place of the individual characterization of aggregates.

This paper presents two automatic methods for visual grading, designed to solve the industrial problem of evaluation of seed lots from the characterization of a representative sample. The sample is thrown in bulk onto a tray placed in a chamber for acquiring color image in a controlled and reproducible manner. Two image processing methods have been developed to separate, and then characterize each seed present in the image. A shape learning is performed on isolated seeds. Collected information is used for the segmentation. The first approach adopted for the segmentation step is based on simple criteria such as regions, edges and normals to the boundary. Marked point processes are used in the second approach, leading to tackle the problem by a technique of energy minimization. In both approaches, an active contour with shape prior is performed to improve the results. A classification is done on shape or color descriptors to evaluate the quality of the sample.

This paper focuses on quality control of mechanical parts in aeronautical context by using a single PTZ camera and the CAD model of the mechanical part. In our approach two attributed graphs are matched using a similarity function. The similarity scores are injected in the edges of a bipartite graph. A best-match-search procedure in bipartite graph guarantees the uniqueness of the match solution. The method achieves excellent performance in tests with synthetic data, including missing elements, displaced elements, size changes, and combination of these cases.

Growing numbers of long tunnels across the globe has increased the need for safety measurements and inspections of tunnels in these days. To avoid serious damages, tunnel inspection is highly recommended at regular intervals of time to find any deformations or cracks at the right time. While following the stringent safety and tunnel accessibility standards, conventional geodetic surveying using techniques of civil engineering and other manual and mechanical methods are time consuming and results in troublesome of routine life. An automatic tunnel inspection by image processing techniques using non rigid registration has been proposed. There are many other image processing methods used for image registration purposes. Most of the processes are operation of images in its spatial domain like finding edges and corners by Harris edge detection method. These methods are quite time consuming and fail for some or other reasons like for blurred or images with noise. Due to use of image features directly by these methods in the process, are known by the group, correlation by image features. The other method is featureless correlation, in which the images are converted into its frequency domain and then correlated with each other. The shift in spatial domain is the same as in frequency domain, but the processing is order faster than in spatial domain. In the proposed method modified normalized phase correlation has been used to find any shift between two images. As pre pre-processing the tunnel images i.e. reference and template are divided into small patches. All these relative patches are registered by the proposed modified normalized phase correlation. By the application of the proposed algorithm we get the pixel movement of the images. And then these pixels shifts are converted to measuring units like mm, cm etc. After the complete process if there is any shift in the tunnel at described points are located.

In this paper, we propose a novel reflection based method to estimate the local orientation of a specular surface. For a calibrated scene with a fixed light band, the band is reflected by the surface to the image plane of a camera. Then the local geometry between the surface and reflected band is estimated. Firstly, in order to find the relationship relying the object position, the object surface orientation and the band reflection, we study the fundamental theory of the geometry between a specular mirror surface and a band source. Then we extend our approach to the spherical surface with arbitrary curvature. Experiments are conducted with mirror surface and spherical surface. Results show that our method is able to obtain the local surface orientation merely by measuring the displacement and the form of the reflection.

Wood singularities detection is a primary step in wood grading enhancement. Our approach is purely machine vision based. The main objective is to compute physical properties like density, modulus of elasticity (MOE) and modulus of rupture (MOR) given wood surface images. Knots are one of the main singularities which directly affect the wood strength. Hence, our target is to detect knots and classify them into transverse and non-transverse ones. Then the Knots Depth Ratio (KDR) is computed based on all found transverse knots. Afterwards, KDR is used for the wood mechanical model improvement. Our technique is based on colour image analysis where the knots are detected by means of contrast intensity transformation and morphological operations. Then KDR computations are based on transverse knots and clear wood densities. Finally, MOE and MOR are computed using KDR images. The accuracy of number of knots found, their locations, MOE and MOR has been validated using a dataset of 252 images. In our dataset, these values were manually calculated. To the best of our knowledge our approach is the first purely machine vision based method to compute KDR, MOE and MOR.

Nowadays, cocoa bean exportation from Piura-Peru is having a positive international market response due to their inherent high quality. Nevertheless, when using subjective techniques for quality assessment, such as the cut test, a wastefulness of grains is generated, additional to a restriction in the selection as well as improvement approaches in earlier stages for optimizing the quality. Thus, in an attempt to standardize the internal features analyzed by the cut test, for instance, crack formation and internal color changes during the fermentation, this research is submitted as an approach which aims to make use of hyperspectral images, with the purpose of having a quick and accurate analysis. Hyperspectral cube size was reduced by using Principal Component Analysis (PCA). The image generated by principal component PC1 provides enough information to clearly distinguish the internal cracks of the cocoa bean, since the zones where these cracks are, have a negative correlation with PC1. The features taken were processed through a fuzzy block, which is able to describe the cocoa bean quality. Three membership functions were defined in the output: unfermented, partly fermented and well fermented, by using trapezoidal-shaped and triangular-shaped functions. A total of twelve rules were propounded. Furthermore, the bisector method was chosen for the defuzzification. Begin the abstract two lines below author names and addresses.

In this paper, we propose an approach to extend the application of anisotropic Gaussian filtering for multi- spectral image denoising. We study the case of images corrupted with additive Gaussian noise and use sparse matrix transform for noise covariance matrix estimation. Specifically we show that if an image has a low local variability, we can make the assumption that in the noisy image, the local variability originates from the noise variance only. We apply the proposed approach for the denoising of multispectral images corrupted by noise and compare the proposed method with some existing methods. Results demonstrate an improvement in the denoising performance.

This paper presents the considerations taken during the conception of a prototype combining two hyperspectral cameras (VNIR and SWIR), dedicated to the characterization of metallic alloys fine-sized particles, coming from end-of-life vehicles and electric and electronic equipment wastes, as well as the calibration steps necessary to obtain quality reflectance data. Classification results obtained on a data-set of 100 metallic fragments, previously characterized with XRF technology, are also presented.

A method to detect certain regions in thermal images of human faces is presented. In this approach, the following steps are necessary to locate the periorbital and the nose regions: First, the face is segmented from the background by thresholding and morphological filtering. Subsequently, a search region within the face, around its center of mass, is evaluated. Automatically computed temperature thresholds are used per subject and image or image sequence to generate binary images, in which the periorbital regions are located by integral projections. Then, the located positions are used to approximate the nose position. It is possible to track features in the located regions. Therefore, these regions are interesting for different applications like human-machine interaction, biometrics and biomedical imaging. The method is easy to implement and does not rely on any training images or templates. Furthermore, the approach saves processing resources due to simple computations and restricted search regions.

The aim of this work is to present a novel approach for automatic recognition of suspicious activities in outdoor perimeter surveillance systems based on infrared video processing. Through the combination of size, speed and appearance based features, like the Center-Symmetric Local Binary Patterns, short-term actions are identified and serve as input, along with user location, for modeling target activities using the theory of Hidden Conditional Random Fields. HCRFs are used to directly link a set of observations to the most appropriate activity label and as such to discriminate high risk activities (e.g. trespassing) from zero risk activities (e.g loitering outside the perimeter). Experimental results demonstrate the effectiveness of our approach in identifying suspicious activities for video surveillance systems.

This paper presents a noise removal and image detail enhancement method that accounts for the limitations on human's perception to effectively visualize high-dynamic-range (HDR) infrared (IR) images. In order to represent real world scenes, IR images use to be represented by a HDR that generally exceeds the working range of common display devices (8 bits). Therefore, an effective HDR mapping without losing the perceptibility of details is needed. To do so, we introduce the use of two guided filters (GF) to generate an accurate base and detail image component. A plausibility mask is also generated from the combination of the linear coefficients that result from each GF; an indicator of the spatial detail that enables to identify those regions that are prominent to present noise in the detail image component. Finally, we filter the working range of the HDR along time to avoid global brightness fluctuations in the final 8 bit data representation, which results from combining both detail and base image components using a local adaptive gamma correction (LAGC). The last has been designed according to the human vision characteristics. The experimental evaluation shows that the proposed approach significantly enhances image details in addition to improving the contrast of the entire image. Finally, the high performance of the proposed approach makes it suitable for real word applications.

We propose a method to classify multiple similar actions which are contained in human behaviors by considering a weak-constrained order of “actions”. The proposed method regards the human behavior as a combination of “action” patterns which have order constrained weakly. In this method, actions are classified by using not only image features but also consistency of transitions between an action and next action. By considering such an action transition, our method can recognize human behavior even if image features of different action are similar to each other. Based on this idea, we have improved the previous HMM-based algorithm effectively. Through some experiments using test image sequences of human behavior appeared in a bathroom, we have confirmed that the average classification success rate is 97 %, which is about 53 % higher than the previous method.

Recently, the research in multispectral face recognition has focused on developing efficient frameworks for improving face recognition performance at close-up distances. However, few studies have investigated the multispectral face images captured at long distance. In fact, great challenges still exist in recognizing human face in images captured at long distance as the image quality might be affected and some important features masked. Therefore, multispectral face recognition tools and algorithms should evolve from close-up distances to long distances. To address these issues, we present in this paper a novel image fusion scheme based on Maximum Ratio Combining algorithm and improve multispectral face recognition at long distance. The proposed method is compared with similar super-resolution method based on the Maximum likelihood algorithm. Simulation results show the efficiency of the proposed approach in term of average variance of detection error.

The 3D chain (acquisition-processing-compression) is, most of the time, sequenced into several steps. Such approaches result into an one-dense acquisition of 3D points. In large scope of applications, the first processing step consists in simplifying the data. In this paper, we propose a coarse to fine acquisition system which permits to obtain simplified data directly from the acquisition. By calculating some complementary information from 2D images, such as 3D normals, multiple homogeneous regions will be segmented and affected to a given primitive class. Contrary to other studies, the whole process is not based on a mesh. The obtained model is simplified directly from the 2D data acquired by a 3D scanner.

In previous works, we have extended the principles of “variable homography”, defined by Zhang and Greenspan, for measuring height of emergent fibers on glass and non-woven fabrics. This method has been defined for working with fabric samples progressing on a conveyor belt. Triggered acquisition of two successive images was needed to perform the 3D measurement. In this work, we have retained advantages of homography variable for measurements along Z axis, but we have reduced acquisitions number to a single one, by developing an acquisition device characterized by 4 lenses placed in front of a single image sensor. The idea is then to obtain four projected sub-images on a single CCD sensor. The device becomes a plenoptic or light field camera, capturing multiple views on the same image sensor. We have adapted the variable homography formulation for this device and we propose a new formulation to calculate a depth with plenoptic cameras. With these results, we have transformed our plenoptic camera in a depth camera and first results given are very promising.

2D projectivity is an invertible mapping to present the perspective imaging of a world plane by projective translation, called homography. Good image feature have to be robust under 2D projectivity caused by any camera movements. In the standard performance evaluation of template matching, many real captured images of many scenes are ordinarily used. However it is not enough to evaluate the robustness under 2D projectivity in detail because the variations of real camera pose and position in the 3D world are limited and the capturing cost is expensive. During the early stage of the template matching development, an easy performance evaluation method is required to examine the behavior. We propose a self-diagnosis method to measure the robustness of local descriptor base template matching between a template image and reference images which are created by projective translation of the template image. We focus on the template matching consisting of a feature point extraction and a local descriptor matching. The proposed method evaluates the spatial accuracy of the feature points and the estimated template positions in the reference images with local descriptor matchings. Four metrics, feature point precision (PP), feature point recall (PR), local descriptor matching precision (MP) and local descriptor matching recall (MR) are introduced to evaluate the performance. The experiment results will be appeared in the final manuscript to show the effectiveness of our method.

3D modeling of scene contents takes an increasing importance for many computer vision based applications. In particular, industrial applications of computer vision require efficient tools for the computation of this 3D information. Routinely, stereo-vision is a powerful technique to obtain the 3D outline of imaged objects from the corresponding 2D images. As a consequence, this approach provides only a poor and partial description of the scene contents. On another hand, for structured light based reconstruction techniques, 3D surfaces of imaged objects can often be computed with high accuracy. However, the resulting active range data in this case lacks to provide data enabling to characterize the object edges. Thus, in order to benefit from the positive points of various acquisition techniques, we introduce in this paper promising approaches, enabling to compute complete 3D reconstruction based on the cooperation of two complementary acquisition and processing techniques, in our case stereoscopic and structured light based methods, providing two 3D data sets describing respectively the outlines and surfaces of the imaged objects. We present, accordingly, the principles of three fusion techniques and their comparison based on evaluation criterions related to the nature of the workpiece and also the type of the tackled application. The proposed fusion methods are relying on geometric characteristics of the workpiece, which favour the quality of the registration. Further, the results obtained demonstrate that the developed approaches are well adapted for 3D modeling of manufactured parts including free-form surfaces and, consequently quality control applications using these 3D reconstructions.

This paper deals with the inspection of an airplane using a Pan-Tilt-Zoom camera mounted on a mobile robot moving around the airplane. We present image processing methods for detection and inspection of four different types of items on the airplane exterior. Our detection approach is focused on the regular shapes such as rounded corner rectangles and ellipses, while inspection relies on clues such as uniformity of isolated image regions, convexity of segmented shapes and periodicity of the image intensity signal. The initial results are promising and demonstrate the feasibility of the envisioned robotic system.

This paper presents a novel scheme for automatic and intelligent 3D digitization using robotic cells. The advantage of our procedure is that it is generic since it is not performed for a specific scanning technology. Moreover, it is not dependent on the methods used to perform the tasks associated with each elementary process. The comparison of results between manual and automatic scanning of complex objects shows that our digitization strategy is very efficient and faster than trained experts. The 3D models of the different objects are obtained with a strongly reduced number of acquisitions while moving efficiently the ranging device.

This paper presents an ultrasonic field simulation software, dedicated to Non Destructive Testing, aiming at interactivity. This work relies on Civa Software semi-analytical model. By restricting its scope to homogeneous isotropic specimens with simple geometry and half-skip modes, an almost completely regular algorithm, well suited to modern hardware, can be derived. The performance of three implementations on multicore SIMD general purpose processors (GPP), manycore accelerators (MIC) and graphical processing units (GPU) over a set of 18 realistic configurations (a standard one plus 17 variations) are presented and analysed. For GPP and the GPU, interactive performances with almost 30 fps have been reached on the standard configuration. This is, to our knowledge, the very first time for a NDT ultrasonic field simulation software.

Surface cracks can be the bellwether of the failure of any component under loading as it indicates the component’s fracture due to stresses and usage. For this reason, crack detection is indispensable for the condition monitoring and quality control of road surfaces. Pavement images have high levels of intensity variation and texture content, hence the crack detection is difficult. Moreover, shallow cracks result in very low contrast image pixels making their detection difficult. For these reasons, studies on pavement crack detection is active even after years of research. In this paper, the fuzzy Hough transform is employed, for the first time to detect cracks on any surface. The contribution of texture pixels to the accumulator array is reduced by using the tiled version of the Hough transform. Precision values of 78% and a recall of 72% are obtaining for an image set obtained from an industrial imaging system containing very low contrast cracking. When only high contrast crack segments are considered the values move to mid to high 90%.

Digital Image Correlation (DIC) based full field strain mapping methodology is used for mapping strain on an aluminum sample subjected to tensile deformation. The local strains on the surface of the specimen are calculated at different strain intervals. Early localization of strain is observed at a total strain of 0.050ɛ; itself, whereas a visually apparent localization of strain is observed at a total strain of 0.088ɛ;. Orientation of the line of fracture (12.0°) is very close to the orientation of locus of strain maxima (11.6°) computed from the strain mapping at 0.063ɛ itself. These results show the efficacy of the DIC based method to predict the location as well as the profile of the fracture, at an early stage.

In this work, we evaluate the relevance of current state of the art algorithms widely employed in the detection of cracks, for the specific context of aerial inspection, which is characterized by image quality degradation. In this study we focus on minimal cost path and on Marked Point Process algorithms, and we test their resilience to motion blur. The results show that the current strategies for defect detection are sensitive to the quality of input images; alternatively, we suggest some improvements based on a-contrario methods that are able to cope with significant motion blur.

In this paper a stable and unsupervised version of FCM algorithm named FCMO is presented. The originality of the proposed FCMO algorithm relies: i) on the usage of an adaptive incremental technique to initialize the class centres that calls into question the intermediate initializations; this technique renders the algorithm stable and deterministic, and the classification results do not vary from a run to another, and ii) on the unsupervised evaluation criteria of the intermediate classification result to estimate the optimal number of classes; this makes the algorithm unsupervised. The efficiency of this optimized version of FCM is shown through some experimental results for its stability and its correct class number estimation.

Melanoma is the deadliest type of skin cancer, yet it is the most treatable kind depending on its early diagnosis. The early prognosis of melanoma is a challenging task for both clinicians and dermatologists. Due to the importance of early diagnosis and in order to assist the dermatologists, we propose an automated framework based on ensemble learning methods and dermoscopy images to differentiate melanoma from dysplastic and benign lesions. The evaluation of our framework on the recent and public dermoscopy benchmark (PH² dataset) indicates the potential of proposed method. Our evaluation, using only global features, revealed that ensembles such as random forest perform better than single learner. Using random forest ensemble and combination of color and texture features, our framework achieved the highest sensitivity of 94% and specificity of 92%.

We present the problem of setting up an intrinsic camera calibration under Scheimpflug condition for an industrial application. We aim to calibrate the Scheimpflug camera using a roughly hand positioned calibration pattern with bundle adjustment technique. The assumptions used by classical calibration methodologies are not valid anymore for cameras undergoing Scheimpflug condition. Therefore, we slightly modify pin-hole model to estimate the Scheimpflug angles. The results are tested on real data sets captured from cameras limited by various industrial constraints, and in the presence of large distortions.

Standard cameras capture only a fraction of the information that is visible to the human visual system. This is specifically true for natural scenes including areas of low and high illumination due to transitions between sunlit and shaded areas. When capturing such a scene, many cameras are unable to store the full Dynamic Range (DR) resulting in low quality video where details are concealed in shadows or washed out by sunlight. The imaging technique that can overcome this problem is called HDR (High Dynamic Range) imaging. This paper describes a complete smart camera built around a standard off-the-shelf LDR (Low Dynamic Range) sensor and a Virtex-6 FPGA board. This smart camera called HDR-ARtiSt (High Dynamic Range Adaptive Real-time Smart camera) is able to produce a real-time HDR live video color stream by recording and combining multiple acquisitions of the same scene while varying the exposure time. This technique appears as one of the most appropriate and cheapest solution to enhance the dynamic range of real-life environments. HDR-ARtiSt embeds real-time multiple captures, HDR processing, data display and transfer of a HDR color video for a full sensor resolution (1280 1024 pixels) at 60 frames per second. The main contributions of this work are: (1) Multiple Exposure Control (MEC) dedicated to the smart image capture with alternating three exposure times that are dynamically evaluated from frame to frame, (2) Multi-streaming Memory Management Unit (MMMU) dedicated to the memory read/write operations of the three parallel video streams, corresponding to the different exposure times, (3) HRD creating by combining the video streams using a specific hardware version of the Devebecs technique, and (4) Global Tone Mapping (GTM) of the HDR scene for display on a standard LCD monitor.

This paper presents a new approach for human action recognition around sleeping with the human body parts locations and the positional relationship between human and sleeping environment. Body parts are estimated from the depth image obtained by a time-of-flight (TOF) sensor using oriented 3D normal vector. Issues in action recognition of sleeping situation are the demand of availability in darkness, and hiding of the human body by duvets. Therefore, the extraction of image features is difficult since color and edge features are obscured by covers. Thus, first in our method, positions of four parts of the body (head, torso, thigh, and lower leg) are estimated by using the shape model of bodily surface constructed by oriented 3D normal vector. This shape model can represent the surface shape of rough body, and is effective in robust posture estimation of the body hidden with duvets. Then, action descriptor is extracted from the position of each body part. The descriptor includes temporal variation of each part of the body and spatial vector of position of the parts and the bed. Furthermore, this paper proposes hierarchical action classes and classifiers to improve the indistinct action classification. Classifiers are composed of two layers, and recognize human action by using the action descriptor. First layer focuses on spatial descriptor and classifies action roughly. Second layer focuses on temporal descriptor and classifies action finely. This approach achieves a robust recognition of obscured human by using the posture information and the hierarchical action recognition.

Calibrating multiple cameras is a fundamental prerequisite for many Computer Vision applications. Typically this involves using a pair of identical synchronized industrial or high-end consumer cameras. This paper considers an application on a pair of low-cost portable cameras with different parameters that are found in smart phones. This paper addresses the issues of acquisition, detection of moving objects, dynamic camera registration and tracking of arbitrary number of targets. The acquisition of data is performed using two standard smart phone cameras and later processed using detections of moving objects in the scene. The registration of cameras onto the same world reference frame is performed using a recently developed method for camera calibration using a disparity space parameterisation and the single-cluster PHD filter.

In this paper, we propose a measurement system to evaluate the swallowing by estimating the movement of the thyroid cartilage. We developed a measurement system based on the vision sensor in order to achieve the noncontact and non-invasive sensor. The movement of the subject’s thyroid cartilage is tracked by the three dimensional information of the surface of the skin measured by the photometric stereo. We constructed a camera system that uses near-IR light sources and three camera sensors. We conformed the effectiveness of the proposed system by experiments.

Tridimensional shape recovery based on Photometric Stereo (PS) recently received a strong improvement due to new mathematical models based on partial differential irradiance equation ratios.1 This modern approach to PS faces more realistic physical effects among which light attenuation and radial light propagation from a point light source. Since the approximation of the surface is performed with single step method, accurate reconstruction is prevented by sensitiveness to noise. In this paper we analyse a well-known parametrization² of the tridimensional surface extending it on any auxiliary convex projection functions. Experiments on synthetic data show preliminary results where more accurate reconstruction can be achieved using more suitable parametrization specially in case of noisy input images.

Among the possible sources of error in 3D-reconstruction using the photometric stereo technique, lighting modeling is often neglected, though it can create a dramatic large-scale bias. In this paper, after recalling the physical definition of a primary Lambertian source (isotropic lightings), we show how to derive a lighting model for several real-world scenarios, including directional lightings, nearby sources and extended planar illuminants. Finally, we show how to calibrate general spatially-varying lightings within a plane, in the case where explicitly modeling the lightings would be tedious.

While composite materials are increasingly used in modern industry, the quality control in terms of vision-based surface inspection remains a challenging task. Due to the often complex and three-dimensional structures, a manual inspection of these components is nearly impossible. We present a photometric stereo sensor system including an industrial robotic arm for positioning the sensor relative to the inspected part. Two approaches are discussed: stop-and-go positioning and continuous positioning. Results are presented on typical defects that appear on various composite material surfaces in the production process.

As Photometric Stereo is a means of measuring the gradient field of a surface, an essential step in the measurement of a surface structure is the reconstruction of a surface from its measured gradient field. Given that the surface normals are subject to noise, straightforward integration does not provide an adequate reconstruction of the surface. In fact, if the noise in the gradient can be considered to be Gaussian, the optimal reconstruction based on maximum likelihood principles is obtained by the method of least-squares. However, since the reconstruction of a surface from its gradient is an inverse problem, it is usually necessary to introduce some form of regularization of the solution. This paper describes and demonstrates the functionality of a library of MATLAB functions for the regularized reconstruction of a surface from its measured gradient field. The library of functions, entitled “Surface Reconstruction from Gradient Fields: grad2Surf Version 1.0” is available at the MATLAB file-exchange

http://www.mathworks.com/matlabcentral/fileexchange/authors/321598

The toolbox is the culmination of a number of papers on the least-squares reconstruction of a surface from its measured gradient field, regularized solutions to the problem, and real-time implementations of the algorithms.^1-4

Photometric stereo is a technique for estimating the 3-D depth of a surface using multiple images taken under different illuminations from the same viewing angle. Most existing models make use of Lambertian reflection and an orthographic camera as underlying assumptions. However, real-world materials often exhibit non-Lambertian effects such as specular highlights and for many applications it is of interest to consider objects close to the camera. In our work, we aim at addressing these issues. Together with a perspective camera we employ a non-Lambertian reflectance model, namely the Blinn-Phong model which is capable to deal with specular reflection. Focusing on the effects of specular highlights, we perform a detailed study of one-dimensional test cases showing important aspects of our method.

A nondestructive inspection method using an infrared detection system is presented in this paper; the system uses a YAG laser as excitation point. The material thermal response to this excitation is processed for the detection of volume defects, this technique integrated into a 3D scanning system allows us to get a 3D scan of the object as well as defects detection.

An infrared system has been developed to monitor transport infrastructures in a standalone configuration. Results obtained on bridges open to traffic allows to retrieve the inner structure of the decks. To complete this study, experiments were carried out over several months to monitor two reinforced concrete beams of 16 m long and 21 T each. Detection of a damaged area over one of the two beams was made by Pulse Phase Thermography approach. Measurements carried out over several months. Finally, conclusion on the robustness of the system is proposed and perspectives are presented.

This study concerns the in-situ determination of the thermal resistance of a building wall. Measurements were performed in the PANISSE platform, which is a residential building with two floors located in the town of Villemomble, at about ten kilometers in the east of Paris. During a renovation, a 6cm-thick external insulating layer was fixed onto the cellular concrete walls of the house. Three methods using IR thermography were used to estimate the thermal resistance of the insulated walls. Results are compared to a standardized method (ISO 9869-1) that uses contact sensors. The comparison is made considering estimated thermal resistance values, measurement uncertainties and measurement duration.

A device allowing heating a liquid to high temperatures during a very short time has been conceived in our laboratory. The goal of this survey is to find the suitable experimental configurations, so that tested material affected by the temperatures coved between 200 and 750°C. This study is achieved to the Solar Furnace of the DGA in Odeillo. The cavity containing the liquid is a thermocouple sleeve (capillary) in Inconel 600. Its extremity is closed tightly by a removable steel plug permitting the tightness after replenishment. An electromagnet associated to a generator of delay permit to make fall the whole after the solar irradiation in liquid nitrogen in order to stop the reaction of "deterioration" of the tested product. According to capillary dimensions and to heating time, the temperature measurement using a pyrometer is not possible. A second possibility is using thermocouple, but it is not easy to join this captor on Inconel 600. Using by infrared camera allows observing the presence or the absence of inflammation during the solar irradiation and the sleeve fall too. The measures of temperatures by thermocouple show a lot of variability. The measures comparison with those by infrared camera shows a phenomenon of "heat well". Several score of tests to the solar furnace have been achieved in different experimental configurations. Nine experimental configurations have been validated, for variable flux of 100 to 500W/cm². The observation by infrared camera permitted to validate the conceived system and to verify the homogeneity of the sleeve heated.

Active thermography is a widely employed technique for parametric identification and non-destructive inspection. This attractive method is based on the observation of thermal waves propagation induced by a periodic heating. For nondestructive testing usual approaches are based on a global heating (a large surface of the inspected material is submitted to thermal excitation). In the following a local approach is investigated: the heated area is small (order of magnitude is one square centimeter) and lateral propagation is studied in order to reveal the defect in the sample. In fact, both modulus (heat wave amplitude) and phase lag (delay) of the measured periodic signal are modified by the defect neighborhood and the search for the most effective area leads to the defect localization. Several results are highlighted in this communication in order to investigate an automated procedure. Temperatures are measured by an infrared camera and analyses of modulus cartography are performed in order to estimate the defect location. In such an aim, the downhill simplex method is implemented in order to converge toward defect location. Illustrations are dedicated to glued structures (two plates separated by a thin glue interface) for which unknown defect is a lack of glue which can be considered as a bubble (air trapped between the lower and the upper plane surface). Automated method attractiveness is established in several configurations.

Infrared thermography is a working technology for over decades, which have been applied mainly in the buildings. We want to move this use to the animal production in order to help us to detect problems of energy efficiency in the facilities preventing, for example, the animal’s welfare. In animal production it is necessary to provide a suitable microclimate according to age and production stage of the animals. This microclimate is achieved in the facilities through the environment modification artificially, providing an appropriate comfort for the animals. Many of the problems detected in farms are related to a poor environmental management and control. This is where infrared thermography becomes an essential diagnostic tool to detect failures in the facilities that will be related with health and performance of the animals. The use of this technology in energy audits for buildings, facilities, etc. is becoming more frequent, enabling the technician to easily detect and assess the temperature and energy losses, and it can be used as a support to draft reports and to transmit the situation to the owner in a visual format. In this way, both will be able to decide what improvements are required. Until now, there was not an appropriate technology with affordable prices and easy to manage enough in order to allow the use of the thermography like a routine tool for the diagnostic of these problems, but currently there are some solutions which are starting to appear on the market to meet the requirements needed by the industry.