This PDF file contains the front matter associated with SPIE Proceedings Volume 6503, including Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

We describe new image analysis developments in support of the U.S. Department of Energy's (DOE) Advanced Gas Reactor (AGR) Fuel Development and Qualification Program. We previously reported a non-iterative, Bayesian approach for locating the boundaries of different particle layers in cross-sectional imagery. That method, however, had to be initialized by manual preprocessing where a user must select two points in each image, one indicating the particle center and the other indicating the first layer interface. Here, we describe a technique designed to eliminate the manual preprocessing and provide full automation. With a low resolution image, we use "EdgeFlow" to approximate the layer boundaries with circular templates. Multiple snakes are initialized to these circles and deformed using a greedy Bayesian strategy that incorporates coupling terms as well as a priori information on the layer thicknesses and relative contrast. We show results indicating the effectiveness of the proposed method.

This paper describes a machine vision dedicated to the dimensional control of power capacitors. The geometry of these pieces is parallepipedic. This system is in keeping with the category of the active vision systems. It is built with two cameras and a LCD projector. This one will be used to determine the deformation of the parallelepiped. The calibration of the system is done with the Faugeras-Toscani's method. The set of points for the calibration is obtained by detecting the corners of the black squares on the 3D pattern. For this, we have used a technique based on the Harris corner detectors. The LCD projector is calibrated by using the last calibration. It is used to determine the deformation of the parallelepiped. The capacitor is held on a rotating table and we examine it from all sides. So we can find all the points of interest of the capacitor. Detected points are stored and will be used to give the dimensional measure. In the beginning of the process, an operator enters the references of the capacitor and the dimensions that he wants to control. All the dimensions of the capacitors produced are stored in a data base. It's easy and quick to check if these measurements are in adequacy with the specifications.

This paper will describe a novel and automated system based on a computer vision approach, for objective evaluation of stain release on cotton fabrics. Digitized color images of the stained fabrics are obtained, and the pixel values in the color and intensity planes of these images are probabilistically modeled as a Gaussian Mixture Model (GMM). Stain detection is posed as a decision theoretic problem, where the null hypothesis corresponds to absence of a stain. The null hypothesis and the alternate hypothesis mathematically translate into a first order GMM and a second order GMM respectively. The parameters of the GMM are estimated using a modified Expectation-Maximization (EM) algorithm. Minimum Description Length (MDL) is then used as the test statistic to decide the verity of the null hypothesis. The stain is then segmented by a decision rule based on the probability map generated by the EM algorithm. The proposed approach was tested on a dataset of 48 fabric images soiled with stains of ketchup, corn oil, mustard, ragu sauce, revlon makeup and grape juice. The decision theoretic part of the algorithm produced a correct detection rate (true positive) of 93% and a false alarm rate of 5% on these set of images.

The objective of this research was to design and fabricate a compact, cost effective multispectral instrument and to collect and analyze spectra for real-time contaminant detection for poultry processing plants. The prototype system developed in this research consisted of a multispectral imaging system, illumination system and an industrial portable computer. The dual-band spectral imaging system developed in this study was a two-port imaging system that consisted of two identical monochrome cameras, optical system and two narrow bandpass filters whose center of the wavelength are 520 and 560 nm with 10 nm FWHM, respectively. A spectral reflectance from a chicken carcass was collected and split in two directions by an optical system including a beamsplitter and lenses, and then two identical collimated lights were filtered by the narrow bandpass filters and delivered to the cameras. Lens distortions and geometric misalignment of the two cameras were mathematically corrected. The prototype system was tested at the real-time processing line and the preliminary results showed that the dual-band spectral imaging system could effectively detect feces and ingesta on the surface of poultry carcass.

In the recent years an experimental hyperspectral multi-camera system based on liquid crystal tunable filters (LCTF) was developed at the Chair for Measurement and Information Technology of the HSU Hamburg. The system allows the acquisition of narrow band image series over the complete spectral range from 400nm - 1800nm. In the range from 400nm - 1100nm CMOS cameras with a dynamic range of up to 22bits are used, allowing radiometrical measurements over 6.5 decades. In the SWIR range an InGaAs camera with a dynamic range of 14 bits is used. The system is intended especially to find out interesting spectral ranges for particular applications.

This paper aims at reviewing the recent published works dealing with industrial applications which rely on Infrared Radiation and of its main variations. Some of its applications are reviewed domain by domain. Distinction between near infrared (visible to 1.1 microns) which enables high temperature measurements, versus classical IR spectrum( SW and LW) range used for low temperature range is done, as well as difference between active and passive thermography.

Thin-film sensors for use in automotive or aeronautic applications must conform to very high quality standards. Due to defects that cannot be addressed by conventional electronic measurements, an accurate optical inspection is imperative to ensure long-term quality aspects of the produced thin-film sensor. In this particular case, resolutions of 1 &mgr;m per pixel are necessary to meet the required high quality standards. Furthermore, it has to be guaranteed that defects are detected robustly with high reliability. In this paper, a new method is proposed that solves the problem of handling local deformations due to production variabilities without having to use computational intensive local image registration operations. The main idea of this method is based on a combination of efficient morphological preprocessing and a multi-step comparison strategy based on logical implication. The main advantage of this approach is that the neighborhood operations that care for the robustness of the image comparison can be computed in advance and stored in a modified reference image. By virtue of this approach, no further neighborhood operations have to be carried out on the acquired test image during inspection time. A systematic, experimental study shows that this method is superior to existing approaches concerning reliability, robustness, and computational efficiency. As a result, the requirements of high-resolution inspection and high-performance throughput while accounting for local deformations are met very well by the implemented inspection system. The work is substantiated with theoretical arguments and a comprehensive analysis of the obtained performance and practical usability in the above-mentioned, challenging industrial environment.

Most of the automation for 3D acquisition concerns objects with simple shape, like mechanical parts. For cultural heritage artefacts, the process is more complex, and it doesn't exist general solution nowadays. This paper presents a method to generate a complete 3D model of cultural heritage artefacts. In a first step, MVC is used to solve the view planning problem. Then, holes remaining in 3D model are detected, and their features are calculated to finish acquisition. Different post-processing are applied on each view to increase quality of the 3D model. This procedure has been tested with simulated scanner, before being implemented on a motion system with five degrees of freedom.

For an automatic tapping of a steel converter and for further processing steps it is necessary to measure the bath level of the steel casting pans (charging ladles). The harsh environment and the properties of molten steel avoid applications of standard bath level detection systems like based on laser, micro wave or super sonic. Instead of these methods within this paper a stereoscopic camera system is introduced which acquires pictures of the molten steel surface. It is possible to detect the bath level without complex equalization of the perspective by geometric calculations. To know which deformations can make the system defective is essential on designing the mechanical bodywork. Simulations with Matlab allow showing the influences of variations to the stereoscopic standard geometry and the boundaries of the system accuracy. Out of these calculations inaccuracies of camera angle appears to be the main parameter for failures. But a developed algorithm is possible to affect that angle after a tapping automatically. Furthermore different parameters have been analyzed to optimize the accuracy of the stereoscopic algorithm. Generally the investigations of the stereo system have shown which specifications including certain tolerances have to be fulfilled by the designer of the mechanical construction to get accurate and reliable results for the automation process.

Machine work pieces with ground, broached or milled surfaces have frequently microtextures consisting of stochastically placed straight tool marks. In this paper we'll exploit the depth data acquired by white-light interferometer for the surface analysis. We present a new algorithm for efficiently extracting extensive groove bands from depth images. The images are treated as a composition of shape component, straight line structures and background. The cylindrical shape component is extracted using robust least squares methods. The outliers are removed by adaptive center weighted median filter. The undefined regions due to sensor failures are interpolated using successive over-relaxation algorithm. An algorithm for the separation of groove bands is introduced. Straight line in three-dimensional space is parameterized and used as primitive for groove separation. After having determined the orientation of the groove band, straight line segment produced with Digital Differential Analyzer can be used by the scanning algorithm for estimating the straight lines. This decomposition enables a separate evaluation of different components of the surface data. The results of the pre-processings and the separation turn out to be fast and robust, which is verified by real depth data.

This paper introduces a method for quasi-motion extraction from a blurred image utilizing edge field analysis (EFA). Exposing a film for a certain time, we can directly photograph the trajectory of the moving object as an edge in a blurred image. As the edge trajectories are not exactly the same but similar to the optical flows, they allow us to treat the edge image as a pseudo-vector field. We define three line integrals in the edge image on closed curve similar to vector analysis. These integrals correspond to three flow primitives of the scene such as the translation, rotation and divergence. As the images, we utilized some images such as the storm, the bottle rocket and a moving object with random patterns. In order to evaluate the proposed method, we conducted the experiments of estimating the eye of the storm, the center of the explosion in terms of bottle rocket, and the centers of the rotation and divergence of the moving object.

This research was focused on a system in which a manipulator with robot vision transfers objects to a mobile robot that moves on a flat floor. In this system, an end effector of a manipulator operates on a plane surface, so only single vision is required. In a robot vision system, many processes are usually needed for vision calibration, and one of them is measurement of camera parameters. We developed a calibration technique that does not explicitly require camera parameters to reduce the number of calibration processes required for our researched system. With this technique, we measured relations between coordinate systems of images and a mobile robot in the moving plane by using a projective transformation framework. We also measured relations between images and the manipulator in an arbitrary plane in the same way. By analyzing the results, we obtained a relation between the mobile robot and the manipulator without explicitly calculating the camera parameters. This means capturing images of the calibration board can be skipped. We tested the calibration technique using an object-transfer task. The results showed the technique has sufficient accuracy to achieve the task.

The subject of this paper is the improvement of measures on imperfect circular forms. Indeed, simple geometric forms have been well studied in image processing. Thus, articles describing circles on a discrete framework are numerous but the case of imperfect geometric forms, in return, is hardly ever deepen. However, it is a classical problem in industrial vision control process to not have a perfect, or perfectly discretized, geometric object due to, notably, manufacturing process, industrial environment (dust, vibrations, objects displacement, etc.), interferences on acquisition chain (electronic noise, lenses imperfections, etc. ). The authors present a comparison of measurement methods of circles characteristics subpixel estimation (center's coordinates and radius) for several distortions (geometric or not). The estimators proposed are classic least mean squares, 3D Hough algorithms and a method combining a Radon transform based estimator and a FitzHugh-Nagumo partial differential equation based active region algorithm. The originality of the method is to furnish a set of geometric envelopes in a single pass from a roughest to a full detailed representation. Moreover, this multiple active region principle also offers interesting electronic implementation possibilities for real time image processing for metrology on production chains.

Generally, for the face image recognition, we must cope with the image shift and image rotation problem. To cope with the image-shifting problem, this research uses one pixel inside the sample image to compare with the around pixels that surrounding the corresponding pixel that inside the unknown image. The "ring rotation invariant transform" technique is used to transfer the geometry feature of the face image to another more salient feature. By this approaching one can obtain more salient geometry feature of the face image. By this more salient geometry feature, one can judge whether or not the sample image and the unknown image are the same image. The "ring rotation invariant transform" technique can solve the image rotation problem. In this research, three different kinds of extracted ring signals are generated. The extracted ring signals are generated by the following ring-circles - ring-radius-31-circle, ring-radius-22-circle, and ringradius- 13-circle. These extracted ring-signals are used to generate the rotation invariant vector magnitude quantities. These rotation invariant vector magnitude quantities are combined as one entity and this entity is saved inside one specific corresponding pixel in the BMP file. By this approach, one pixel will possess more geometry-features of the face images. The obtained entity of the combined signals of one specific pixel inside the sample image will be compared to the entities of the combined signals of the entire pixels located in the corresponding radius-6-cake in the unknown image. By this comparison, one can find the most-matching point of the geometry-feature of the pixels between the sample image and the unknown image.

The visual revolution triggered by the commercial application of digital image capturing devices generates the need for new miniaturized and cheap optical imaging systems and cameras. However, in imaging we observe a permanent miniaturization of elements but always similar optical principles are applied which have been known to the optical designers for many decades. With the newly gained spectrum of technological capabilities in micro- optics such as photolithography it is time to exploit completely new imaging principles such as for instance the microlens array imaging. In this paper we present an overview of our latest developments on: the technology and image processing of the artificial apposition compound eye, a rotating artificial apposition compound eye column for panoramic vision, an artificial apposition compound eye on a curved basis and an ultra-short, large object-size microscope. All the systems have a total track of below or only a few mm in common, while at the same time having an optical performance comparable to that of the conventional exemplars, e.g. a resolution of 50LP/mm over a field of 4.5mm for the large object-size microscope.

Today, printing products which must meet highest quality standards, e.g., banknotes, stamps, or vouchers, are automatically checked by optical inspection systems. Typically, the examination of fine details of the print or security features demands images taken from various perspectives, with different spectral sensitivity (visible, infrared, ultraviolet), and with high resolution. Consequently, the inspection system is equipped with several cameras and has to cope with an enormous data rate to be processed in real-time. Hence, it is desirable to move image processing tasks into the camera to reduce the amount of data which has to be transferred to the (central) image processing system. The idea is to transfer relevant information only, i.e., features of the image instead of the raw image data from the sensor. These features are then further processed. In this paper a color line-scan camera for line rates up to 100 kHz is presented. The camera is based on a commercial CMOS (complementary metal oxide semiconductor) area image sensor and a field programmable gate array (FPGA). It implements extraction of image features which are well suited to detect print flaws like blotches of ink, color smears, splashes, spots and scratches. The camera design and several image processing methods implemented on the FPGA are described, including flat field correction, compensation of geometric distortions, color transformation, as well as decimation and neighborhood operations.

Illumination with correct intensity and direction can help solving even very complicated inspection and measurement problems by hiding unnecessary information and by enhancing the contrast where it is needed. The 'wave and look' approach used widely in the vision industry requires knowledge about different illumination techniques and the measurement problem. One illumination is not optimal for all tasks performed using one vision system and so even more complex illumination combinations can be expected, particularly on small-batch production lines. We describe approaches for producing flexible light sources for machine vision purposes. Our devices allow one to change the scene illumination by software. Often only one programmable light source is needed per one vision system despite very different illumination needs. We also describe an automated tuning approach that controls programmable light sources on-line to correct the image seen by the sensor to maximize the illumination uniformness, contrasts, or even the part measurement robustness. The paper also shows the results of test cases and discusses ways to further improve them.

Confocal imaging is primarily based on the use of apertures in the detection path to provide the acquired three-dimensional images with satisfactory contrast and resolution. For many years, it has become an important mode of imaging in microscopy. In biotechnology and related industries, this technique has powerful abilities of biomedical inspection and material detection with high spatial resolution, and furthermore it can combine with fluorescence microscopy to get more useful information. The objective of this paper is first to present a generalized theoretical framework for confocal imaging systems, and then efficiently to design and implement such systems with satisfactory imaging resolutions. In our approach, a theoretical review for confocal imaging is given to investigate this technique from theory to practice. Also, computer simulations are performed to analyze the imaging performance with varying optomechanical conditions. For instance, the effects of stray light on the microscopic systems are examined using the simulations. In this paper, a modified optomechanical structure for the imaging process is proposed to reduce the undesired effects. From the simulation results, it appears that the modified structure highly improves the system signal-to-noise ratio. Furthermore, the imaging resolution is improved through the investigation on the tolerance of fabrication and assembly of the optical components. In the experiments, it is found that the imaging resolution of the proposed system is less sensitive than that of common microscopes, to the position deviations arising from installations of the optical components, such as those from the pinhole and the objective lens.

In this paper, upon the background of driving assistance on highway, we propose a real-time vehicle detection and tracking algorithm based on traffic scene analysis. We describe a general traffic scene analysis framework for vehicle detection and tracking based on roadside detection at first. On that basis, we present a new object detection algorithm via fusion of global classifier and part-based classifier and a vehicle detection algorithm integrating classifying confidence and local shadow. The local shadow is obtained by detecting the Maximally Stable Extremal Regions (MSER) using a multi-resolution strategy. Finally, we test our algorithm on several video sequence captured from highway and suburban roads. The test results show high efficiency and robustness when coping with environment transition, illumination variation and vehicle orientation change.

Specular surfaces inspection remains a delicate task within the automatic control of products made by plastic plating. These objects are of very varied shape and their surface is highly reflective acting like a mirror. This paper presents steps to follow in order to detect geometric aspect surface defects on objects made by plastic plating. The projection of a binary fringes pattern is used and enables to reveal the defects near the transition between a dark fringe and a bright fringe. Indeed, the surface imperfections provoke important light rays deviations. By moving this dynamic lighting, and thanks to a saturated camera, the system brings an aspect image where the defects appear very contrasted on a dark background. A simple image processing algorithm is then applied leading to a very efficient segmentation. To obtain such resulting images, the translation step, the duty cycle and also the number of images are constraint. This article finally shows how to adjust these parameters according to the various sizes of defect and to the objects shape.

Ornamental stone products are commercially classified on the market according to several factors related both to intrinsic lythologic characteristics and to their visible pictorial attributes. Sometimes these latter aspects prevail in quality criteria definition and assessment. Pictorial attributes are in any case also influenced by the performed working actions and the utilized tools selected to realize the final stone manufactured product. Stone surface finishing is a critical task because it can contribute to enhance certain aesthetic features of the stone itself. The study was addressed to develop an innovative set of methodologies and techniques able to quantify the aesthetic quality level of stone products taking into account both the physical and the aesthetical characteristics of the stones. In particular, the degree of polishing of the stone surfaces and the presence of defects have been evaluated, applying digital image processing strategies. Morphological and color parameters have been extracted developing specific software architectures. Results showed as the proposed approaches allow to quantify the degree of polishing and to identify surface defects related to the intrinsic characteristics of the stone and/or the performed working actions.