The ability to identify and quantify changes in the microstructure of metal alloys is valuable in metal cutting and shaping applications. For example, certain metals, after being cryogenically and electrically treated, have shown large increases in their tool life when used in manufacturing cutting and shaping processes. However, the mechanisms of microstructure changes in alloys under various treatments, which cause them to behave differently, are not yet fully understood. The changes are currently evaluated in a semi-quantitative manner by visual inspection of images of the microstructure. This research applies pattern recognition technology to quantitatively measure the changes in microstructure and to validate the initial assertion of increased tool life under certain treatments. Heterogeneous images of aluminum and tungsten carbide of various categories were analyzed using a process including background correction, adaptive thresholding, edge detection and other algorithms for automated analysis of microstructures. The algorithms are robust across a variety of operating conditions. This research not only facilitates better understanding of the effects of electric and cryogenic treatment of these materials, but also their impact on tooling and metal-cutting processes. Future work will focus on the application of statistical methods for analyzing images of more complex metal alloys.

This paper addresses the automated detection of line features in large industrial inspection images. The manual examination of these images is labor-intensive and causes undesired delay of inspection results. Hence, it is desirable to automatically detect certain features of interest. In this paper we are concerned with the detection of vertical or slanted line features that appear at unpredictable intervals across the image. The line features may appear distorted due to shortcomings of the sensor and operator conditions. Line features are modeled as a pair of smoothed step edges of opposite polarity that are in close proximity, and two operators are used to detect them. The individual operator-outputs are combined in a non-linear fashion to form the line-feature response. The line features are then obtained by following the ridge of the line-feature response. In experiments on four datasets, over 98.8% of line features are correctly detected, with a low false-positive rate. Experiments also show that the approach works well in the presence of considerable noise due to poor operating conditions or sensor failure.

This paper is oriented to study techniques to improve the precision of the systems for wear measurement of contact wire in the railways. The problematic of wear measurement characterized by some important determining factors like rate of sampling and auscultation conditions is studied in detail. The different solutions to resolve the problematic successfully are examined. Issues related to image acquisition and image processing are discussed. Type of illumination and sensors employed, image processing hardware and image processing algorithms are some topics studied. Once analyzed each one factor which have influence on the precision of the measurement system, there are proposed an assembly of solutions that allow to optimize the conditions under which the inspection can be carried out. Part of the development exposed in this paper is the result of an investigation work in a project in collaboration with the Electronic Division of Engineering of the Escuela Técnica Superior de Ingenieros Industriales of the Universidad Politécnica de Madrid and the business RENFE (Spanish Railways Company) that has culminated in a system for measurement of contact wire (MEDES) which actually is been used on the laboratories coach of RENFE and the SNCF (France) The system is European patent.

This paper aims at developing a novel defect detection algorithm for the semiconductor assembly process by image analysis of a single captured image, without reference to another image during inspection. The integrated circuit (IC) pattern is usually periodic and regular. Therefore, we can implement a classification scheme whereby the regular pattern in the die image is classified as the acceptable circuit pattern and the die defect can be modeled as irregularity on the image. The detection of irregularity in image is thus equivalent to the detection of die defect. We propose a method where the defect detection algorithm first segments the die image into different regions according to the circuit pattern by a set of morphological segmentations with different structuring element sizes. Then, a feature vector, which consists of many image attributes, is calculated for each segmented region. Lastly, the defective region is extracted by the feature vector classification.

In the last decade, the accessibility of inexpensive and powerful computers has allowed true digital holography to be used for industrial inspection using microscopy. This technique allows capturing a complex image of a scene (i.e. containing magnitude and phase), and reconstructing the phase and magnitude information. Digital holograms give a new dimension to texture analysis since the topology information can be used as an additional way to extract features. This new technique can be used to extend previous work on image segmentation of patterned wafers for defect detection. This paper presents a combination of features obtained using Gabor filtering on different complex images. The combination enables to cope with the intensity variations occurring during the holography and provides final results which are independent from the selected training samples.

As the electronic industry advances rapidly, the shrunk dimension of the device leads to more stringent requirement on process control and quality assurance. For instance, the tiny size of the solder bumps grown on wafers for direct die-to-die bonding pose great challenge to the inspection of the bumps’ 3D quality. Traditional pattern projection method of recovering 3D is about projecting a light pattern to the inspected surface and imaging the illuminated surface from one or more points of view. However, image saturation and the specular nature of the bump surface are issues. This paper proposes a new 3D reconstruction mechanism for inspecting the surface of such wafer bumps. It is still based upon the light pattern projection framework, but uses the Ronchi pattern - a pattern that contrasts with the traditionally used gray level one. With the use of a parallel or point light source in combination with a binary grating, it allows a discrete pattern to be projected onto the inspected surface. As the projected pattern is binary, the image information is binary as well. With such a bright-or-dark world for each image position, the above-mentioned difficult issues are avoided. Preliminary study shows that the mechanism holds promises that existing approaches do not.

In this paper, we propose a coarse-to-fine image comparison algorithm based on Hausdorff distance for PCB inspection. The Hausdorff distance can be used in a geometrics-based inspection framework for comparing binary edge maps extracted from the inspection images. To use the Hausdorff distance for image alignment, we need to compute the edge map from the input image as the first step. In some cases, one may use directed Hausdorff distance as a similarity measure in order to reduce the computational cost during the image alignment. Moreover, a modified version of directed Hausdorff distance is employed to enforce robustness against random noises introduced by edge detection. The search for the optimal alignment by minimizing the associated Hausdorff distance is accomplished by an efficient multi-resolutional downhill simplex search algorithm. In addition to the image alignment, we also apply a modified Hausdorff distance to detect defects in PCB. In our inspection system, we apply the partial Hausdorff distance in a local circuit window to reduce the inspection area dramatically, thus making it very efficient for PCB inspection. Experimental results on some PCB inspection examples are shown to demonstrate the accuracy and efficiency of the proposed Hausdorff-distance based inspection system.

Oak Ridge National Laboratory (ORNL) has begun the development of a program for the manufacturing and characterizing fuel pellets for use in advanced nuclear reactors. To achieve high reliability it is necessary to characterize the pellets during production runs. In this paper we present a simple TRISO Particle Counting And Sizing Tool (TP-CAST) that performs dual measurements of counting and size estimation for particles at rates up to 200 per second. The TP-CAST is based on a laser with line-generation optics and a PC-based data acquisition and analysis system. The instrument can measure 1000 micron pellets with a standard deviation of approximately 11 microns and with counting errors less than 0.075%. Our paper discusses the signal modeling, algorithms for size estimation, system design, and experimental results of the prototype TP-CAST system assembled at ORNL.

Precision rolled strips are often intermediate products in the manufacturing of blades. In such cases the shape and size of these strips are essential to the functionality and quality of the blade and cutting workpiece. Although precision strips are normally produced in heavily automated rolling mills, their size and shape are still inspected manually with profile gauges and microscopes. In this paper we present a measurement setup with multiple light-sectioning systems, which is suitable for the inspection of all sides of a profiled strip. It consists of three measurement heads, which are used to inspect the upper side, the lower side and the back of the blade. The heads are calibrated individually; the focus of the work here is to determine the relative position and orientation of the heads with respect to each other. The first approach has been developed to reference two or more measurement heads. The calculation of the required transformations is based on the rotation of a suitable target. Due to the small depth of field, the location of the rotation axis must be pre-adjusted very precisely. To improve the accuracy and to simplify the process, a second referencing method was developed. The required target was manufactured by means of a 5-axis high speed milling machine and features a thickness tolerance of less than 1 micron. Both the referencing method and target are presented. Additionally, we demonstrate the all-side inspection of a blade. It will be shown that the approaches allow a robust and flexible referencing of multiple measurement heads to each other.

The overall quality of a fabric is dependent on a number of factors. Among these is the fabric’s tendency to wrinkle after home laundering - referred to as smoothness. Wrinkle grading is a subjective process involving human graders who compare fabric samples to replicas, representing various degrees of wrinkling. This process is also operator dependent, expensive, and it lacks the ability to adequately describe the many subtle differences that exist between grades. Therefore, the textile industry needs an automated system that can describe wrinkles on a fabric surface in an objective and repeatable manner. In this paper, we describe a computer vision system developed in a previous work and examine the effectiveness of new features extracted from the wavelet domain independent mixture model and a landform classification technique. Shown to be useful in texture classification, features from the wavelet domain independent mixture model are measured based on the two-population characteristic of the wavelet domain. The second technique uses topographical analysis methods originally developed for geographical landform classification that have been successfully applied to digital elevation models of the Earth’s surface. These new measurements, representing quantitative descriptions of the surface of a fabric in both the frequency and spatial domains, are compared to the existing industry grading standard using a fuzzy classifier. Results show a good correlation with technicians’ grades.

Conventional smoke detection systems currently installed onboard aircraft are often subject to high rates of false alarms. Under current procedures, whenever an alarm is issued the pilot is obliged to release fire extinguishers and to divert to the nearest airport. Aircraft diversions are costly and dangerous in some situations. A reliable detection system that minimizes false-alarm rate and allows continuous monitoring of cargo compartments is highly desirable. A video-based system has been recently developed by Goodrich Corporation to address this problem. The Cargo Fire Verification System (CFVS) is a multi camera system designed to provide live stream video to the cockpit crew and to perform hotspot, fire, and smoke detection in aircraft cargo bays. In addition to video frames, the CFVS uses other sensor readings to discriminate between genuine events such as fire or smoke and nuisance alarms such as fog or dust. A Mamdani-type fuzzy inference engine is developed to provide approximate reasoning for decision making. In one implementation, Gaussian membership functions for frame intensity-based features, relative humidity, and temperature are constructed using experimental data to form the system inference engine. The CFVS performed better than conventional aircraft smoke detectors in all standardized tests.

Deflectometry has proven to be a very precise and reliable technique for the detection and measurement of bumps, dents, waviness and scratches on specular surfaces. Phase shifted fringe patterns are successively reflected at the surface and the spatial distortion of these reflected patterns is observed with a camera to extract information about the shape of the surface. Up to now, deflectometry could not be used for diffuse reflecting surfaces, because specular reflection does not occur. With the system developed at our institute it is now possible to inspect even diffuse reflecting surfaces like unpolished metal or plastics using the deflectometric measuring principle. Hereby the fact is exploited that a surface becomes specular when the reflected light has sufficiently large wavelength compared to the surface roughness. For the diffuse surfaces mentioned above the adequate range of the electromagnetic spectrum is far-infrared. In our approach a reflected infrared pattern is observed with a thermal camera. By analyzing four images of the phase shifted pattern an image is calculated, which contains information about local surface curvature. The presented method has been successfully tested for the inspection of the diffuse surfaces of unpainted car body parts.

This work gives some practical, simulated and calculated design parameters for the detection of voids inside the material with active thermography for different void geometry, orientation and depths. Main goal is to find the limitations of detectability for different materials and voids, to help designers for test systems with: a quick estimation of the feasibility and to find the necessary camera parameters. The methods used (algebraic, numeric and practical) to find these values will be described. For inline applications the so called square pulse technique is easy to automate and needs less power from the source, because energy can be brought into the probe for a longer time span. Further its strength (in relation to flash pulse technique) is to find voids deeper below the surface. Therefore all of the calculations and practical verifications will be done only with square pulse. The finite difference calculations are used to get a quick approximation for the dimensioning parameters. Some hints how to work with this method and how to prevent errors will be given in this paper. Practical tests with artificial probes and known void properties will be done with some of the parameters to verify the calculated values.

The motivation for this work arises from the need for a contactless extension measurement system to obtain material properties of refractories at elevated temperatures of up to 1400°C. There is a lack of tactile probes capable of performing reliable measurements at this temperature range. Presently, the deformation of the specimen is estimated by measuring the movement of the testing machine. To determine material properties such as the modulus of elasticity more accurate information about the real deformation is required. This work presents an optical arrangement and mechanical setup for image acquisition at high temperatures and at a high degree of thermal radiation. Furthermore, an efficient and stable image processing algorithm based on edge detection and line fitting using singular value decomposition is presented. Subsequently the sources of measurement errors are determined and examined. The accuracy of the measurement result is estimated.

The environmental conditions at hot strip mills are characterized by hot objects (up to 1100 °C), dust and dense steam. Conventional methods for characterizing product parameters fail in these unfavorable measurement conditions. An alternative is optical measurement technique. It is non-contact and can be positioned in a safe location in the plant. One parameter of interest at hot strip mills that is to be observed is the geometry of the transfer bar. During the reduction of thickness of the stock in several reversing passes a deviation of the straightness, a so-called camber, can occur. This distortion can cause disturbances at the following steps of production, e.g. at the finishing mill. The value and direction of the camber is determined by means of a CCD-matrix camera and a segmented capture of the shape. The unique character of the measuring system discussed in this paper lies in the use of a single camera system to meet the requirements in contrast to other systems using two or more cameras. In the rolling process of the roughing mill with varying speed additional phenomena like lateral shift and rarely twist can occur, which make the measurement more difficult. Lateral movements of the strip are taken into consideration by checking single edges twice, the twist can be neglected. Finally the results of the camber values are statistically evaluated. One significant result shows the influence of the arrangement of stocks in the pusher-type furnaces on the camber. The results of the measurement are helpful for understanding the rolling process in more detail and even to take measures to avoid cambers and improve quality and process stability.

This paper will describe the distributed industrial inline application “broken roll detection”, which is placed in a really harsh industrial environment, with all aspects from the sensing base to algorithm, implementation and technology. In a seamless steel tube production the pipe shells produced in the punch bench are running through many roller stands (3-roll system) to get the final dimension. If one of the rolls is broken, structural voids near the surface are the consequence. So finding the structure voids on the tube means to find broken rolls. Since pipe shells are hot (approximately 900°C) after passing the rolls, temperature distribution on its surface is different when voids happen. This gives a good base for detecting such voids by watching the surface temperature by sensing the radiation at wavelengths from 0,7 to 1.1μm, which means that standard line scan cameras (3 x 2048 pixels, 10kHz line rate) can be used. Images of up to 600MB are the result for each imaged pipe shell. Evaluation of image data is done stepwise (in a pipeline) and on a separate channel for each camera with the objective to reduce data at each step. Images are detruncated, position-normalized, filtered, segmented and converted into object-descriptions that are sent to another PC for evaluating periodic occurrences. Once found such a periodic occurrence, the system signalizes it to the production line to stop the machine and repair the broken roll.

Many optical inspection systems today can capture surface slope information directly or indirectly. For these systems, it is possible to perform a 3-D surface reconstruction which converts surface slopes to surface heights. Since the slope information obtained in such systems tend to be noisy and sometimes heavily quantized, a noise-tolerant reconstruction method is needed. We used a simple bayes reconstruction method to improve noise tolerance, and multi-resolution processing to improve the speed of calculations. For each resolution level, the surface slopes between pixels are first calculated from the original surface slopes. Then the height reconstruction for this resolution level is calculated by solving the linear equations that relate relative heights of each point and its related surface slopes. This is done through a Bayesian method which makes it easier to incorporate prior knowledge about height ranges and noise levels. The reconstructions are done for a small window of pixels at a time for each resolution level to make the linear equations manageable. The relative height solutions from all resolution levels are then combined to generate the final height map. This method has been used in optical inspection applications where slope data are quite noisy.

Sheet metal strain analysis is an important tool to ensure products are manufactured within necessary tolerances. A common technique involves electrochemically etching a dark grid pattern of known size onto the flat sheet metal surface and then deforming the sheet. The change in the grid pattern after deformation can be used to calculate surface strain. The computer vision problem is to accurately detect the intersections of the grid pattern. To investigate this problem, a stereo camera system was designed and attached to a bridge style coordinate measurement machine. The stereo head consists of two high resolution monochrome CCD cameras mounted on a Renishaw PH10 motorized probe head that can be articulated into numerous, repeatable, preset positions. Stereo head calibration was achieved using Zhang’s technique with a planar target. Each probe position was calibrated using a global point set registration method to link coordinate systems. A novel approach to segmenting the grid pattern into squares involving region merging and watersheds is described. Grid intersections are determined to sub pixel accuracy and matched between images using a correlation based scheme. The accuracy of the system and experimental results are provided.

We propose a new application of « Shape from Polarization » method to reconstruct surface shapes of specular metallic objects. Studying the polarization state of the reflected light is very useful to get information on the surface normals. After reflection, an unpolarized light wave becomes partially linearly polarized. Such a wave, can be described by its three parameters: intensity, degree of polarization, and angle of polarization. By using the refractive index of the surface, a relationship between the degree of polarization and the reflection angle can be established. Unfortunately, the relation commonly used for dielectrics, cannot be applied since the refractive index of metallic surfaces is complex. To get a similar relation, we apply a usual approximation valid in the visible region. The Fresnel reflectance model can also provide a relationship between the angle of polarization and the incidence plane orientation. Thus, the reflection angle and the incidence plane orientation give the surface normals. The shape is finally computed by integrating the normals with a relaxation algorithm. Applications on metallic objects made by stamping and polishing are also described, and show the efficiency of our system to discriminate shape defects. Future works will consist in integrating the system into an automatic process of defects detection.

This paper presents a digital image processing system for the measurement of the volumetric tensile properties of polymer materials. The standard measurement setup with a single camera system focussing the face of the samples under test is described. An algorithm for fitting parallel and/or orthogonal lines based on singular value decomposition is explained in detail. Applications where higher resolution and accuracy is desired, are addressed by extending the standard measurement setup to two camera systems, whose images can be acquired simultaneously. The non-overlapping fields of view require determination of the relative location of the camera systems. A calibration target and procedure is presented to solve this referencing task. Three-dimensional specimens are subject to motion in testing machines with flexible jaws. An extended hardware setup with two line-projecting lasers is introduced, that enables the measurement of the specimen motion and with this the elimination of the associated errors. In a calibration procedure, the relative locations of the lasers and the camera are determined. The image processing task is extended to calculate the three-dimensional specimen location in each step of the sequence based on the results of the calibration. As a consequence, deformations are determined on the specimen surface under the assumption that the material remains planar. Finally, a number of measurement results gained with the video-extensometer system at real experiments in a testing laboratory are presented and different evaluation algorithms are compared.

In order to efficiently digitize the 3-D shape of objects in applications such as quality control, reverse engineering and inspection, a flexible system based on a portable range sensor coupled to an optical tracking device has been developed. The hand-held range sensor can be moved freely in space without the constraint on motion imposed by a translation or rotation system. An optical tracking device synchronized with the sensor is used to compute the sensor's orientation and location in real-time. The optical tracking and the integration of a 3-D range sensor with a positioning device for manual scanning constitute the two major challenges of the proposed system. This paper first presents the range sensor and the optical tracking system. The main aspects including the calibration approach and the sensor integration are described in the following sections. An error analysis has been conducted to predict the expected results. Finally, experimental results are presented to validate the overall system.

We describe in this paper a stereoscopic system based on a multispectral camera and a projector. To be used, this system must be calibrated. This starts by a geometrical calibration of the stereoscopic set using a weak calibration. It is also necessary to know the spectral response of each element in the acquisition chain, from the projector to the camera. Then, image acquisition can begin. To acquire a multi-spectral image, we have just to use the projector to send a luminous pattern on the scene. The projection of the pattern in the image is detected and labeled since the projector was characterized during the calibration step. Finally, we can obtain the 3D position of the different parts of the luminous pattern on the scene by using triangulation. Moreover, a spectral reflectance can be associated to each of them. The colorimetric accuracy obtained by a multispectral camera is totally improved compared with a color camera.

Flatness inspection is a fundamental issue in the quality control performed during the manufacturing of steel strips. The quality requirements for such products have been increasing for the last years, and nowadays new flatness measurement techniques are needed to fullfill this requirements. This paper introduces the concept of flatness and the most common flatness metrics used in the industry. Our work focuses on the application of the well-known laser ranging techniques in the design, construction and test of a flatness measurement and inspection system capable of aquiring the true shape of a steel strip in real time and calculating its flatness indexes so they can be used in the quality inspection of the steel products. For testing the system, a steel strip simulator has been constructed that allows us to generate any possible flatness defect with a known magnitude and measure it with our flatness measurement system. The agreement between the real magnitude of the defects and the measured ones is better than 4.5 I-Units in the range of 0 to 100 I-Units. The system prototype has been installed in the conditioning line of Aceralia (steel manufacturer) to test the proposed solution under real industrial conditions.

In Computer vision there are a lot of applications that are based on the 3D vision. For example, object modeling for reverse engineering in manufacturing, map building, industrial inspection and so on. However, the surface acquired by most part of sensors only represents a part of the object. To solve this problem, several images of the same object are acquired in different positions. After that, all views are transformed to the same coordinate system. This process is known as Range Image Registration and it is the goal of this work. This work surveys the most common registration methods. These kinds of methods are used to reconstruct 3D complete models of objects. Moreover, a classification of the registration methods is presented. This classification is based on the accuracy of the results. In this survey the principal methods are classified and commented. In order to compare them, experimental results are performed using synthetic and real data. The quality of some results indicates that the result of the registration can be used to compare a real object with the 3D models of them. It can be used in manufacturing process in order to inspect the quality of the produced objects.

An algorithm of fault diagnosis is proposed for vision module on the intelligent agent. The basic theory to generate the expected area is of the fundamental matrix, which is an important concept of epipolar geometry. The proposed method can be performed in real time to judge whether the system is in good health. Compared with the conventional methods, our method doesn’t need additional sensors and requires less CPU load. Experiment shows that the method is reasonable.

In this paper, a smart algorithm is presented to accurately recognize and measure three transparent objects, GRIN lens (circle), thin-film filter (square), and epoxy (random) which are essential parts of a DWDM core. After trial-and-errors, a specially arranged illumination and a black color metal coated 10x10 tray, which was applied to perfectly absorb the light traveled to the bottom surface of the core, were chosen to simplify the algorithm. During the image capturing procedure, the first image was focus on the filter’s top surface and the second images were taken after the stage was moved 0.7 mm lower onto the GRIN lens’ top surface. Threshold and filtering were then used to process images. Blob analysis, edge detection and Angle-of-sight (AOS) signature were utilized to find the exact centers of the lens and the filter. During the preliminary tests, the successful recognition rate was 95% (34 out of 36 pcs.). The two failures were due to the severely damaged edges of the filter. As shown in the results, the computer can easily perform the vision functions just like human beings as long as the human thinking had been considered in the algorithm. As mentioned above, a special light source can simplify the challenging task such as the separation of several transparent objects. By teaching the machine to analyze the AOS signature, it can recognize an object naturally. This is one kind of approaches for human to solve problems. The computer programs were initially developed using the script language provided by the Matrox’s Inspection 3 and ported into the Microsoft Visual Basic (VB) with the Matrox’s MIL image library for VB later. The entire hardware was designed as a plug-in module and the software used the object oriented (OO) concept which can be straightforwardly employed by other programmers.

In this paper, we suggest to use a 2-D LOG filter to inspect Cluster Mura defects on the FOS images of LCDs, either for round-type Cluster Mura defects or rectangular-type Cluster Mura defects. With the 2-D LOG filter, the optimal threshold is analyzed with the SEMU formula. Also, we propose a curvature test approach to detect V-Band Mura defects. In the curvature test approach, a 1-D LOG filter is used to achieve the curve with the smooth curvature tendency. With this estimated curve, V-Band Mura defects could be detected easily. The FOS surface reconstruction verifies this detection approach in a reasonable way. Either for the Cluster Mura detection approach or for the V-Band Mura detection approach, the simulation results demonstrate the LOG filters is very useful in the development of detection algorithms for automatic optical inspection of Mura-like defects.

Robust segmentation of complex images is a challenging problem. Performance with traditional use of statistical information, such as intensity, first and second derivatives, and local intensity histograms, is often degraded severely by noise. Good results with model-based segmentation approaches are generally sensitive to the precise initialization of the model within the image to be segmented. The basic idea of this paper is to integrate a set of image attributes into a single, unified framework, such that their complementary information facilitates reliable and robust decisions in image segmentation. This paper proposes a new vector clustering method, called Attributed Vector Quantization (Attributed VQ), for this purpose. The new method can be considered as a variation of the existing weighted vector quantization method, but uses a different weighting scheme from the traditional method, which is clearly motivated by the complex image segmentation problem. Furthermore, the vector quantization process intrinsically produces hierarchically organized structural information that can characterize the pattern of the object to be segmented. We demonstrate the new method with industrial images as the example. We demonstrate, preliminarily, that this approach shows promise in its application to training and recognition in industrial vision systems by requiring minimal user interaction during training, and by leveraging its basis in vector quantization to reduce sensitivity to noise and other anomalies during recognition.

Trash content of raw cotton is a critical quality attribute. Therefore, accurate trash assessment is crucial for evaluating cotton’s processing and market value. Current technologies, including gravimetric and surface scanning methods, suffer from various limitations. Furthermore, worldwide, the most commonly used method is still human grading. One of the best alternatives to the aforementioned approaches is 2D x-ray imaging since it allows a thorough analysis of contaminants in a very precise and quick manner. The segmentation of trash particles in 2D transmission images is difficult since the background cotton is not uniform. Furthermore, there is considerable overlap between the gray levels of trash and cotton. We dealt with this problem by characterizing and identifying the background cotton via scale-space filtering, followed by a “background normalization” process that removes the background cotton, while leaving the trash particles intact. Furthermore, we have successfully employed stereo x-ray vision for recovering the depth information of the piled trash in controlled samples. Finally, the proposed technique was tested on 280 cotton radiographs-with various trash levels-and the results compared favorably to the existing systems of cotton trash evaluation. Given that the approach described here provides the trash mass in real-time, when realized, it will have a wide-spread impact on the cotton industry.

A complete and practical range image sensor development is presented in this paper: from the mathematical modeling to the shape reconstruction. This scanner aims to be integrated in a larger collaborative project. The nal goal is to provide a framework to allow easy comparisons of ancient wooden items by historians. Motivations and expected results are clearly stated in accordance to nancial and easy-to-use constraints. In order to alleviate the calibration process a new calibrating pattern is proposed. The pattern allow both calibration of camera and projector. The method is validated with experimental results. Experimental results are given for the calibration process and the range image acquisition. These results have been performed on both real and synthetic data, which allows us to comment quantitative performances as well as qualitative ones. They are quite encouraging and satisfactory.