In this paper an infrared system for short-circuit detection in the copper electrorefining process is presented. The system consists of an IR-camera, a computer, radiomodems and software including the developed algorithm to process a thermal image. The basic component of the proposed system is an infrared camera mounted in an air-conditioned protection unit on a moving crane. The video output of the infrared camera is connected to the input of a framegrabber card in a computer. The framegrabber card with software captures a thermal image of the electrolytic cell, then processes it to locate the hot spots (short-circuits in a cell). The inspection results are transferred directly by radio link to the control room to be printed and further processed. The system presented in this paper is a prototype that has been tested for several months. The test results indicate that strong short-circuits can be detected with the proposed system as reliably as with a currently used manual method (gaussmeter). The advantages of the proposed system are easier and faster measurements (all cathodes in a cell can be measured remotely at the same time) and possibility to gather new process information.
Many industrial manufacturing processes are not well understood and are treated as `black art' with few experts able to control the process and ensure product quality. However, modern manufacturing companies are finding it increasingly difficult to compete in the global marketplace without better process understanding and control. Automated inspection systems for general manufacturing have become more feasible through technical advances, primarily in sensor and computing technology. However, these systems have been used almost exclusively for the detection and subsequent removal of well defined, discrete defects from the product; thus guaranteeing high quality for the customer. This paper describes a larger opportunity to affect operations by employing web inspection techniques to dynamically analyze manufacturing conditions rather than just detecting the presence of defective material. One can then keep the process under better control, thereby eliminating defects, ensuring product quality, and optimizing manufacturing time on the production line. Specific image and data processing techniques will be illustrated including product uniformity metrics, automatic determination of thresholds for blob analysis, and localization of repeating defects within production data. The benefit of these techniques will be demonstrated through `real-world' examples of web-based manufactured products.
The paper industry has long had a need to better understand and control its papermaking process upstream, specifically at the wet end in the forming section of a paper machine. A vision-based system is under development that addresses this need by automatically measuring and interpreting the pertinent paper web parameters at the wet end in real time. The wet-end characterization of the paper web by a vision system involves a 4D measurement of the slurry in real time. These measurements include the 2D spatial information, the intensity profile, and the depth profile. This paper describes the real-time depth profile measurement system for the high-speed moving slurry. A laser line-based measurement method is used with a high-speed programmable camera to directly measure slurry height. The camera is programmed with a profile algorithm, producing depth data at fast sampling rates. Analysis and experimentation have been conducted to optimize the system for the characteristics of the slurry and laser line image. On-line experimental results are presented.
Spurious reflection from shiny surfaces is one of the major problems in machine vision inspection systems based on triangulation of structured light. To investigate this problem, a non-contact real-time 3D machine vision inspection prototype is described. Its accuracy is +/- 2 micrometers using a structured laser light sheet imaged by a machine vision camera. To avoid the spurious reflections from shiny surfaces, two methods are investigated. In the first method, the object is scanned off-line, finding the 3- space locations of the object's surface points using standard triangulation algorithms. Next the object's surfaces are reconstructed, breaking the surfaces into various simple feature primitives such as planes, cylinders, cones and spheres. The influence of spurious reflections from shiny surfaces is eliminated by deleting those points that do not fall on the reconstructed surfaces. In the second method, a 3D datacloud density method is employed. The datacloud density of a 3D point is the number of measured points in the close vicinity of that point. Since a spurious point's datacloud density is generally lower than a correct point lying on the part's surface, the spurious points can be identified and deleted.
This paper deals with an on line defects detection system. This system is used for controlling surfaces reflecting as mirror. The originality of this work is the lighting used and the filter used to process the image. Developed lighting is a structured one. It is composed of a succession of luminous and dark stripes. Using this lighting, defects appear clearly on the images. Images obtained with this specific lighting are particular and very contrasted, for these reason an original filter, very easy to implement has been developed to process them. This filter looks like a morphological filter. Conception of the operator has been done in accordance to the knowledge of the shape of the defects to be detected. One of the important constraint of the system is the processing time, which must be very short to work on a standard PC at the flow rate of the industrial process.
This work describes a real-time continuous broiler weighting system based on machine vision, used for size sort planning in a process plant. We demonstrate that this technology can be used successfully as an alternative to classical electromechanical carcasses weighting system. A digitized silhouette image of the carcass hung by its feet is divided in six regions: the legs, the wings, the neck and the breast. Once the parts are separated, their individual areas are measured and used in a polynomial equation that predicts the overall bird weight. A sample of 1400 birds were collected, labeled and weighted in a precision scale, in different days and different hours. We found an error standard deviation of 78 grams for broilers weighing from 750 to 2100 grams. The morphological image processing algorithms demonstrated to be robust to extract the individual parts of the carcass.
The image reference approach is very popular in industrial inspection due to its generality for different inspection tasks. Unfortunately, this approach is sensitive to illumination variations. A novel illumination compensation algorithm is proposed in this paper for correcting smooth intensity variations due to illumination changes. By using the proposed algorithm as a preprocessing step in the image reference based inspection or localization, we can make the image inspection or localization algorithm robust against spatially smooth illumination changes. This technique is very useful to achieve a reliable automated visual inspection system under different illumination conditions. The proposed illumination compensation algorithm is based on the assumption that the underlining image reflectance function is approximately piecewise constant and the image irradiance function is spatially smooth. Reliable gradient constraints on the smooth irradiance function are computed and selected from the image brightness function by using a local uniformity test. Two surface fitting algorithms are presented to recover the smooth image irradiance function from the selected reliable gradient constraints. One is a polynomial surface fitting algorithm and the other is a spline surface fitting algorithm. The spline surface fitting formulation leads to solving a large linear system, which is accomplished by an efficient preconditioned conjugate gradient algorithm. Once the image irradiance function is estimated, the spatial intensity inhomogeneities can be easily compensated. Some experimental results are shown to demonstrate the usefulness of the proposed algorithm.
A real-time pilot system for defect detection and classification of web textile fabric is presented in this paper. The general hardware and software platform, developed for solving this problem, is presented and a powerful novel method for defect detection is proposed. This method gives good results in the detection of low contrast defects under real industrial conditions, where the presence of many types of noise is an inevitable phenomenon. For the defect classification an artificial neural network, trained by using a back-propagation algorithm, is implemented. Using a reduced number of possible defect classes, the system gives consistent and repeatable results with sufficient speed.
This paper presents a method of searching segmentation parameters which has been developed for an industrial study. The problem consists in the detection of four types of defects on textured industrial parts: smooth surfaces, bumps, lacks of material and hollow knocked surfaces. The lighting system used in this application is not described in this paper but is presented in a previous study concerning the characterization of lighting.
We have developed an automated optical passive alignment technique for planar lightwave circuit (PLC) modules. Our technique is based on aligning a laser diode (LD) on a PLC module, and can be used to create an optical network unit. The PLC module we propose consists of a LD and a photodiode, which are mounted on the surface of the PLC platform without a lens. Because these elements send light directly to the waveguide on the PLC platform and receive light from the waveguide, a precise alignment technique is required. We therefore developed the mirror image alignment method in order to automatically align the LD on the PLC with extreme accuracy. The method is effective regardless of the position of the LD and the thickness of the solder. The mirror image method uses the markers on the PLC and their images, which are reflected on the front wall of the LD. The achieved accuracy for positioning was within 1 micron in the lateral direction and within 0.5 degrees in the rotational direction. These systems are now being used at a Fujitsu factory.
Vision systems for color measurement should be designed with the basic principles of colorimetry in mind. A lack of understanding that a vision based color measurement system could fall if it ignores the basic principles of colorimetry is the main reason for some earlier color vision system failures. The purpose of this paper is to clarify how the color measurement principles have to be applied to vision systems and how the electro-optical design features of colorimeters have to be modified in order to implement them for vision systems. Despite their considerable advantages color vision systems are not capable of solving all color and appearance problems. Not understanding the limitations of this new technology is another reason for the slow progress of color vision systems. Sometimes color vision systems have been used for solving problems which, for a variety of reasons cannot be successfully resolved with them. A critical overview of the major areas of applications for color vision systems will be presented.
This paper presents a method for the automatic evaluation of the appearance of seam puckers (SP) on suits. Presently, evaluations are done by inspectors who compare standard photographs to test samples. We regard the evaluation as pattern recognition, using the fractal dimensions as the features of SP. The first difficult point of automatic evaluation is that the gray levels of SP are often confused with the gray levels of material's texture. We solved the problem by distinguishing the SP from the texture using the concept of variance. For images containing SP we apply the contrast transform. By these processes, confusion is avoided. The second point is the calculation of fractal dimensions. In order to make it easy to calculate fractal dimensions, we make a curve representing the property of SP. From the curve fractal dimension is calculated. Twenty suits were used as test patterns for the evaluation experiment, and a good result was obtained. We also made an evaluation system using Daubechies' wavelet and compared it with the present system. The evaluation results obtained by the system using the fractal dimensions showed a better result than that of the wavelet feature.
The high-speed production of textiles with complicated printed patterns presents a difficult problem for a colorimetric measurement system. Accurate assessment of product quality requires a repeatable measurement using a standard color space, such as CIELAB, and the use of a perceptually based color difference formula, e.g. (Delta) ECMC color difference formula. Image based color sensors used for on-line measurement are not colorimetric by nature and require a non-linear transformation of the component colors based on the spectral properties of the incident illumination, imaging sensor, and the actual textile color. This research and development effort describes a benchtop, proof-of-principle system that implements a projection onto convex sets (POCS) algorithm for mapping component color measurements to standard tristimulus values and incorporates structural and color based segmentation for improved precision and accuracy. The POCS algorithm consists of determining the closed convex sets that describe the constraints on the reconstruction of the true tristimulus values based on the measured imperfect values. We show that using a simulated D65 standard illuminant, commercial filters and a CCD camera, accurate (under perceptibility limits) per-region based (Delta) ECMC values can be measured on real textile samples.
In case of robot vision, most important problem is the processing speed of acquiring and analyzing images are less than the speed of execution of the robot. In an actual robot color vision system, it is considered that the system should be processed at real time. We guessed this problem might be solved using by the bicolor analysis technique. We have been testing a system which we hope will give us insight to the properties of bicolor vision. The experiment is used the red channel of a color CCD camera and an image from a monochromatic camera to duplicate McCann's theory. To mix the two signals together, the mono image is copied into each of the red, green and blue memory banks of the image processing board and then added the red image to the red bank. On the contrary, pure color images, red, green and blue components are obtained from the original bicolor images in the novel color system after the scaling factor is added to each RGB image. Our search for a bicolor robot vision system was entirely successful.
We present in this paper the methods that were used to detect appearance defects on turkey carcasses on a slaughter line. In order to compare two segmentation methods, we also propose means to estimate the quality of detection in an objective way. We show in this paper that the use of color histograms turn out to be an efficient solution to such a problem, when the color gamut is confined and the color classes are close to each other. Also, the methods presented in this paper can easily be used for similar problems and are suitable for hi-speed color detection.
The needs for accurate and efficient object localization prevail in many industrial applications, such as automated visual inspection and factory automation. Image reference approach is very popular in automatic visual inspection due to its general applicability to a variety of inspection tasks. However, it requires very precise alignment of the inspection pattern in the image. To achieve very precise pattern alignment, traditional template matching is extremely time-consuming when the search space is large. In this paper, we present a new FLASH (Fast Localization with Advanced Search Hierarchy) algorithm for fast and accurate object localization in a large search space. This object localization algorithm is very useful for applications in automated visual inspection and pick-and-place systems for automatic factory assembly. It is based on the assumption that the surrounding regions of the pattern within the search range are always fixed, which is valid for most industrial inspection applications. The FLASH algorithm comprises a hierarchical nearest-neighbor search algorithm and an optical-flow based energy minimization algorithm. The hierarchical nearest-neighbor search algorithm produces a rough estimate of the transformation parameters for the initial guess of the iterative optical-flow based energy minimization algorithm, which provides very accurate estimation results and associated confidence measures. Experimental results demonstrate the accuracy and efficiency of the proposed FLASH algorithm.
Geometrical description of object surface for 3D range image provides useful invariant features if the necessary first- and second-order partial derivatives are accurately estimated. Two main methods are used in this study to approximate the partial derivatives. One method uses convolutions between derivative operators and range data. The other method is based on local least squares surface fitting. In this report a least squares surface fitting method based on a set of orthogonal polynomials is introduced to extract the desired 3D surface geometrical features. Given a point and its adjacent points, a local least squares surface model using discrete orthogonal polynomials is obtained. The partial derivatives along with the curvatures of the local surface are then computed according to the polynomial functions. The polynomial parameters are also used as the 3D features to describe the local surface. The experiments show that all of the partial derivatives and the polynomial parameters can be used as surface description features for classification and recognition of 3D objects in range images.
The ability to manage large image databases has been a topic of growing research over the past several years. Imagery is being generated and maintained for a large variety of applications including remote sensing, art galleries, architectural and engineering design, geographic information systems, weather forecasting, medical diagnostics, and law enforcement. Content-based image retrieval (CBIR) represents a promising and cutting-edge technology that is being developed to address these needs. To date, little work has been accomplished to apply these technologies to the manufacturing environment. Imagery collected from manufacturing processes have unique characteristics that can be used in developing a manufacturing-specific CBIR approach. For example, a product image typically has an expected structure that can be characterized in terms of its redundancy, texture, geometry, or a mixture of these. Defect objects in product imagery share a number of common traits across product types and imaging modalities as well. For example, defects tend to be contiguous, randomly textured, irregularly shaped, and they disrupt the background and the expected pattern. We will present the initial results of the development of a new capability for manufacturing-specific CBIR that addresses defect analysis, product quality control, and process understanding in the manufacturing environment. Image data from the semiconductor-manufacturing environment will be presented.
Deconstructing an image based upon it parts poses a challenge to image analysis that may be solved using adaptive algorithms. The presence of occlusion or image rotation makes template matching difficult. Image segmentation techniques can be used to discriminate between objects via feature synthesis using deformable templates. This paper describes modifications to existing techniques commonly used to do real-time image segmentation for efficient hardware implementation. Edge detection and edge direction finding techniques may be used within the context of deformable templates for real-time automatic target recognition and tracking.
In this paper, we present several methods to improve the processing speed of an optoelectronic morphological industrial inspection processor, which uses an incoherent correlator as its optical hardware and the extensive complementary encoding morphological hit-or-miss transform as its detection algorithm. The first method is using a multi-channel correlation scheme, in which four database images are processed simultaneously so that the LCD panel needs only update 25 times for a set of 100 images, for instance. The second method is using a postprocessing method for the optical correlation output plane. An absolute difference measurement algorithm is applied for measuring the similarity of the optical correlation resultant images, and then the similarity between the tested and the reference image can be deduced. By this method, the complicated preprocessing procedure including the extensive complementary encoding can be simplified because we can directly deal with the gray-scale images. The third method is using a photorefractive correlator instead of the incoherent correlator so that their is not the update rate limitation of the LCD panel, which was a main problem facing any optoelectronic hybrid system.
Automated inspection of manufactured parts is a challenging problem and a crucial issue for any production process. When concerned with dimensional tolerancing, CAD models are a highly desirable reference, as they describe precisely the parts under inspection. However, the time needed to identify the right model in a large CAD database is far too long for on-line inspection because of the huge amount of data to be analyzed. In this paper, we propose to use a vision-based representation for the parts which is much less accurate than CAD descriptions, but offers very low storage requirements, thus speeding up the identification stage. We discuss the inspection system proposed and report our preliminary results in converting CAD models to their simplified representation.
This paper describes a vergence control algorithm for a 3D stereo recovery system. This work has been developed within framework of the project ROBTET. This project has the purpose of designing a Teleoperated Robotic System for live power lines maintenance. The tasks involved suppose the automatic calculation of path for standard tasks, collision detection to avoid electrical shocks, force feedback and accurate visual data, and the generation of collision free real paths. To accomplish these tasks the system needs an exact model of the environment that is acquired through an active stereoscopic head. A cooperative algorithm using vergence and stereo correlation is shown. The proposed system is carried out through an algorithm based on the phase correlation, trying to keep the vergence on the interest object. The sharp vergence changes produced by the variation of the interest objects are controlled through an estimation of the depth distance generated by a stereo correspondence system. In some elements of the scene, those aligned with the epipolar plane, large errors in the depth estimation as well as in the phase correlation, are produced. To minimize these errors a laser lighting system is used to help fixation, assuring an adequate vergence and depth extraction.
Technological advances provide now the opportunity to automate the pavement distress assessment. This paper deals with an approach for achieving an automatic vision system for road surface classification. Road surfaces are composed of aggregates, which have a particular grain size distribution and a mortar matrix. From various physical properties and visual aspects, four road families are generated. We present here a tool using a pyramidal process with the assumption that regions or objects in an image rise up because of their uniform texture. Note that the aim is not to compute another statistical parameter but to include usual criteria in our method. In fact, the road surface classification uses a multiresolution cooccurrence matrix and a hierarchical process through an original intensity pyramid, where a father pixel takes the minimum gray level value of its directly linked children pixels. More precisely, only matrix diagonal is taken into account and analyzed along the pyramidal structure, which allows the classification to be made.
This paper presents a new system for inspecting 3D manufactured machine parts with high accuracy. The system focuses on two main aspects: a definition of specific tools for inspection and 3D measurement and high flexibility for feature selection. As a result, a novel system for inspecting objects with 3D characteristics has been developed. The input information is a complete knowledge of the inspection workbench setting (elements, characteristics and resolution ranges) and a CAD model of the part to be inspected. Using an interactive interface, the user may define the features to inspect and the precision required for each one. Some of the operations the system performs are dimensional control with subpixel accuracy, surface inspection and object edge finish. Based on the CAD model and the features to inspect the system automatically designs an inspection planning responsible for managing the difference resources involved in the inspection process. Provided that the aim of the system is to obtain the greatest possible accuracy, a great effort has been done in the area of mechanical devices and camera calibration. Also, in order to quantify the goodness of the results obtained, an uncertainty propagation strategy has been carried out throughout the measurement process.
This paper describes a machine vision, automatic inspection algorithm of the printings of soft drink cans by image processing analysis. There are two new techniques employed in this procedure to make the automatic inspection possible: (1) barcode referencing: we develop a fast barcode detection algorithm, such that when the cans pass through the image- taking area in the inspection lines with uncertain directions, we use barcode location as the reference point, (2) 2D matching: we connect the multiple view-angle images of the whole 3D cans' surfaces, then with artificial 2D images, we adjust the matching process for a flexible system inspection resolution requirements for the quality control decision making. This process inspects 3D cans with true color information and can easily replace different cans for inspection.
Traditionally, automated optical inspection techniques have been sufficient to find defects on wafers. However, defects smaller than 0.25 (mu) simply cannot be resolved by light optics due to optical limitation. KLA-Tencor SEMSpec is an advanced scanning electron-beam wafer inspection system that can detect defects down to 0.1 (mu) . The electron beam is scanned over the specimen surface, causing secondary electrons to be emitted from the surface where the beam impinges. These emitted secondary electrons are detected and the resulting image is formed so that conventional image processing techniques can also be applied. In the inspection sequence, fundamental parameters such as pixel size, defect detection sensitivity threshold, signal averaging, and cluster distance can be adjusted.
For OEMs, system integrators and end-users of machine vision requiring highly accurate and robust pattern finding tools capable of precisely locating patterns despite normal process variations, Imaging Technology Incorporated offers its SmARTTM (Smart Alignment and Registration Tool) Search. SmART Search is an extremely accurate and robust pattern locating tool featuring the industry's first and only Training WizardTM--a time-saving utility that takes the guesswork and uncertainty out of the pattern training process. With state-of-the-art GeoSearchTM-assist, SmART Search enables manufacturers of vision-automated equipment to build more robust machines that will automatically adapt to changes in object appearance due to normal process variations.