Search

Building of three-dimensional models is an important topic in computer vision. Range finders only let to reconstruct a partial view of the object. However, in most part of applications a full reconstruction is required. Many authors have proposed several techniques to register 3D surfaces from multiple views. In this paper, a survey of the most common techniques is presented. Furthermore experimental results are performed, and a 3D model is obtained.

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.

Time delay and integration (TDI) imaging was first applied in the early 1970's to aerial photoreconnaissance as an electronic replacement for photographic film. Since that time, TDI has been applied elsewhere infrequently with bare printed circuit board inspection and document digitizing as the only well-known recent applications. Now, TDI has been effectively brought to bear on an ubiquitous problem, that of inspection of moving webs. Although other imaging techniques have been and are being used for web inspection, none shows the broad adaptability and ease of use of TDI-based imaging. Herein is a review of the TDI technique, a specific implementation of TDI in a unique web-imaging-compatible camera and a comparison of the TDI technique with other current technologies.

Many industrial inspection such visual quality control of metal laminates require billions of operations per second. Analog CNN array computer arises as an alternative to traditional digital processors, capable of make in a single chip Tera equivalent operations per second. A 4096 analog CNN processor array is able to perform complex space-time image analysis, being much faster than a camera-computer system in continuous inspection applications. Both chips have been implemented in CMOS technology and they are managed by a 32- bit high-performance low-cost micro-controller that closes the pan, tilt, lighting, focus and zoom loops required in the implementation of the active vision strategies. Several convolution masks for the Cellular Processors has been selected to detect particular changes in the texture, size, direction or orientation of the image entities, reprogramming 'on the fly' the pixel resolution or shape when necessary. Laboratory results present these Cellular Processors and multiple resolution imager circuits as a promising architecture for visual inspection of industrial processes in real time.© (2000) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

In this paper, we present several methods to improve the processing speed of an optoelectronic morphologicalindustrial inspection processor, which uses an incoherent correlator as its optical hardware and the extensive complementaryencoding morphological hit-or-miss transform as its detection algorithm. The first method is using a multi-channelcorrelation scheme, in which four database images are processed simultaneously so that the LCD panel needs only update 25times for a set of 100 images, for instance. The second method is using a postprocessing method for the optical correlationoutput plane. An absolute difference measurement (ADM) algorithm is applied for measuring the similarity of the opticalcorrelation resultant images, and then the similarity between the tested and the reference image can be deduced. By thismethod, the complicated preprocessing procedure including the extensive complementary encoding can be simplifiedbecause we can directly deal with the gray-scale images. The third method is using a photorefractive correlator instead of theincoherent correlator so that there is no the update rate limitation of the LCD panel, which was a main problem facing anyoptoelectronic hybrid system.Keywords: industrial inspection, morphological, processing speed, multi-channel correlation, postprocessing, similaritymeasure, photorefractive correlators

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.

The past decade has seen the development of multispectral and hyperspectral imaging spectrometers for use in remote sensing applications in the aerospace business. Correspondingly, advanced electronic imaging techniques have been exploited for use in industrial inspection and manufacturing process control. TRW has been involved in hyperspectral imaging since 1989 for use in remote sensing of earth resources and has developed many instruments and related technologies which can easily be re-applied to unique industrial inspection applications. These instruments operate in the visible, near-infrared and short-wave infrared wavebands covering the range from 0.4 microns to 2.5 microns depending on the application. The exploitation of hyperspectral imagers for remote sensing has shown the power of spectral imaging for typing and discrimination tasks, which can be readily applied to industrial applications. In this paper we explain the relevant fundamentals of hyperspectral imaging and how it can be exploited for industrial inspection and process control tasks, particularly those that require color or spectral typing and discrimination. The associated technologies used to perform measurements and reduce the data also are described.

The paper describes a low-cost industrial inspection system (MAQVIS) developed by the authors and based on a personal computer equipped with a simple frame grabber and a digital input/output controller. Originally developed under a contract with an industrial partner for inspecting text and symbols printed on electronic components, MAQVIS can be applied to many inspection tasks since it is based on a generic template matching algorithm. The task of inspection is broken down into two steps; first, location of the symbols that must be verified in the image and second, verification of the symbols using criteria previously defined by the operator during the set-up phase. The first step is important in many applications where it is impossible, due to lack of mechanical precision, to guarantee that the object being inspected (or the symbols printed upon it) will always appear in the same position relative to the camera. The second step produces a pass/fail result for each symbol and finally for the component itself. MAQVIS can be programmed to display failed symbols on-screen and indicate the reason for the failure, or simply divert the offending component to the reject bin. Both the location and the verification steps are performed on gray-level images using algorithms specifically developed for this application. Statistical information about the number of failed components and the failure rate of individual symbols is generated and can be used to trace problems in the manufacturing process. The paper describes the system in general, and specifically the template matching algorithm used to locate symbols in the image and also the verification algorithm that validates each symbol as it is found. Results of tests performed using the system on the production line of an electronic components manufacturer are also presented.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

In this paper a High-speed Opto-electric Image Processing Unit (HSOEIPU) based on an incoherent correlator is constructed, which can do two main processing works. One is digital filtering; the other is object recognition. Using HSOEIPU a Lapalace filter is realized to extract the edge characteristic from the images. Then the preprocessed images are sent into the object recognition unit for identifying the important areas or targets. A novel template-matching method is proposed for gray-tone image recognition. A changeable cycle-encoding method is introduced to realize the fuzzy relation pixel-matching on a correlator structure. The system has good fault-tolerance ability for rotation distortion, gaussian noise disturbance or information losing. In the experiments photographs of workpieces are sent into HSOEIPU for processing.

This paper presents an overview of past and current activities in the field of computer vision at Joanneum Research. Joanneum Research is a non-profit research organization covering mostof its revenues from contract research with industrial partners and international agencies in the area of space research (e.g., the European Space Agency). The focus on industrial real-time image processing started with the development of a 2D inspection system for wooden surfaces, and has expanded to the development of a 3D vision system for spacecraft navigation and elevation modeling.In this presentation the major projects in this area of research will be outlined. 1 . INTRODUCTION Our work in real-time image processing typically deals with inspection problems for industrialautomation. The systems developed can handle data rates of 4Mpixel/sec and up using scaleable multiprocessor systems. Projects completed include automatic inspection of • Profiled boards and textured surfaces;• bearing shells;• wet blue hides

This paper describes an image-based inspection system for surface defects, implemented on a transputer system with a high- performance transfer bus to provide fast access to large blocks of data. Data-level parallelism (the image pixel data is partitioned horizontally into slices) and task-level parallelism (the algorithms themselves can be parallelized) are utilized. Surface defects and anomalies are detected by a texture-based segmentation procedure. In the training phase the objects of interest are marked and all feature vectors implemented in the system are computed. The system uses simple statistical features, features calculated from the cooccurrence matrix, features based on texture spectrum and the fractal dimension. The time critical run phase is realized on a parallel computer. The implementation is done fully in software, allowing flexibility in the use of features and window sizes for pixel descriptors and freedom in the use of various classifier-methods. The processing speed of the segmentation system is easily scalable with the number of processing units. The performance of the system is demonstrated on an industrial visual inspection task, the recognition of surface defects of aluminium cast workpieces, where a connectionist classifier is used for pixel classification.

High development costs of machine vision systems can be reduced by designing more general algorithms that can be used in a wide range of applications, by using modular system architectures in which new algorithms and more computational power can be added easily, depending on the needs of the given applications, and by using powerful tools for the design of, for example, new algorithms, hardware, software, optics, and illumination. This paper overviews the related research in progress at the University of Oulu. The topics to be discussed include hybrid computer architecture for machine vision, a color segmentation algorithm based on hierarchical connected components analysis, an interactive tool for performance analysis of parallel vision programs, and an object-oriented image processing library.

The suitability of a number of systems architectures for use in industrial inspection machines is briefly reviewed. Two different pipe-line processors, a (linear) array processor, an array of transputers, a concurrent array of (non-interconnected) von Neumann machines are all discussed. Where possible, measured execution speeds are given. The article concludes that no one system yet available is able to fulfil anything approaching a significant proportion of all inspection requirements. However, a clear view of the short-term future emerges, in which a combination of certain current ideas is perceived as providing a basis for a far more cost-effective 'general purpose' image processing system for industrial inspection than we have yet enjoyed.© (1987) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A real-time Vander Lugt correlator for the automatic recognition, location and inspection of objects from direct imaging or TV camera input is described. Techniques to cope with changes due to scale and rotation are discussed. These include: (i) Image rotation with dove prisms. (ii) Multiple matched spatial filters by parallel addressing. (iii) Selective spatial filtering. Example results from an inspection task are presented.© (1986) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This paper describes an automated visual inspection system for complex industrial purposes. The system is composed of four main interacting divisions. These represent knowledge, processing modules, data structures and the master program. The knowledge division is composed of a model and a set of rules. The model describes geometrical, positional and relational properties of the items under test. Rules are further classified into representational, procedural and control rules. Representational rules are a set of design rules that should be followed by items under inspection. Procedural rules are the specific steps that should be executed to ensure the validity of representational rules. Control rules are respon-sible for the sequence of different levels of system activities. Processing modules are simple image processing, morphological and pattern recognition procedures, each assigned a particular job. They share data at different levels of representation. The master program, supervized by the procedural and control rules, chooses the appropriate processing module to continuously update the data or create a new pictorial structure in order to achieve specific goals. The seperation between the rules and the master program makes it a learning system in the sense that it can be taught new models, design rules and inspection algorithms. Thus, it can be easily adapted to different applications.

Now that image processing techniques applied to industrial inspection have become well established, vision systems are being used in manufacturing in increasing numbers. This paper examines various approaches to multiple vision systems and introduces the concept of a distributed vision system. This concept is illustrated by describing how a number of compact inspection modules can be connected to a single supervisory computer to form a distributed vision network.

The image processing of contour type fringe patterns to extract coordinate information is well understood. This paper discusses the further processing of this information in order to determine the fundamental geometrical nature of the underlying surface. Particular reference is made to the use of such analysis techniques for industrial inspection of surface form.

Mapvision is a photogrammetric machine vision system developed by the Technical Research Centre of Finland primarily for industrial inspection and assembly control applications. The system consists of four simultaneously operating solid state cameras and an especially programmed photogrammetric microprocessor. The three-dimensional measuring accuracy is better than 1:5,000. The Mapvision is an automated system enabling for example the dimensional feed-back and interaction between computer aided design and manufacturing phases.

Images from coherent laser sources are severely degraded by a mixture of speckle and incoherent additive noise. In digital holography, Bayesian approaches reduce the incoherent noise, but prior information are needed about the noise statistics. On the other hand, non-Bayesian techniques presents the shortcomings of resolution loss or very complex acquisition systems, required to record multiple uncorrelated holograms to be averaged. Here we propose a fast non- Bayesian method which performs a numerical synthesis of a moving diffuser in order to reduce the noise. The method does not depend on prior knowledge of the noise statistics and the proposed technique is one-shot, as only one single hologram capture is required. Indeed, starting from a single acquisition multiple uncorrelated reconstructions are provided by random sparse resampling masks, which can be incoherently averaged. Experiments show a significant improvement, close to the theoretical bound. Noteworthy, this is achieved while preserving the resolution of the unprocessed image.

International audience; Accurate estimation of displacement between successive images is a significant topic in the measurement of in-plane vibrations of microscopic objects such as micro-actuators. Actual subpixel motion estimation algorithms require the interpolation of interpixel values which undesirably increases the overall complexity and data flow and deteriorates estimation accuracy. Methods that do not use interpolation for achieving subpixel accuracy are scarcer in the literature. One approach for subpixel movement estimation without interpolation is based on phase correlation algorithm. This algorithm estimates the relative shift between two image blocks by means of a normalized cross-correlation function computed in the 2-D spatial Fourier domain. Indeed, the method is based on the Fourier shift theorem. The cross power spectrum of two images, containing subpixel shifts, is a polyphase decomposition of a Dirac delta function. By estimating the sum of polyphase components one can then determine subpixel shifts along each axis. Phase correlation is the state of the art for interpolation-free subpixel shift measurement between two frames, but this method is strictly limited to subpixel shifts. So, we have implemented this method using a standard optical microscope in order to observe subpixel translations with high spatial resolution measurements (down to 1 nm in the best cases). In this paper, we propose an application of this method to characterize the vibration mode shapes of a small silicon beam used in near-field acoustic microscopy. Harmonic movements of a few tens of nanometers are measured and presented.

New applications for ultra-violet imaging are emerging in the fields of drug discovery and industrial inspection. High throughput is critical for these applications where millions of drug combinations are analyzed in secondary screenings or high rate inspection of small feature sizes over large areas is required. Sarnoff demonstrated in1990 a back illuminated, 1024 X 1024, 18 um pixel, split-frame-transfer device running at > 150 frames per second with high sensitivity in the visible spectrum. Sarnoff designed, fabricated and delivered cameras based on these CCDs and is now extending this technology to devices with higher pixel counts and higher frame rates through CCD architectural enhancements. The high sensitivities obtained in the visible spectrum are being pushed into the deep UV to support these new medical and industrial inspection applications. Sarnoff has achieved measured quantum efficiencies > 55% at 193 nm, rising to 65% at 300 nm, and remaining almost constant out to 750 nm. Optimization of the sensitivity is being pursued to tailor the quantum efficiency for particular wavelengths. Characteristics of these high frame rate CCDs and cameras will be described and results will be presented demonstrating high UV sensitivity down to 150 nm.

Heursistics for programming real-time color recognition hardware, controlled by a symbolic (Prolog) processor, are described. The paper addresses the engineering aspects of color recognition and its application to such tasks as industrial inspection and sorting. The techniques discussed are heuristic, not algorithmic; they are not particularly well suited to those applications where there are subtle color changes. However, in numerous experiments, they have been found to produce useful results, in a wide range of important industrial inspection and sorting tasks, where relatively coarse color discrimination is sufficient. The application of these techniques to the inspection of standard multi-color packaging materials is highlighted.

Computerized systems for industrial inspection of simple forms or limited variables have been wide spread. However, the automated inspection of complex patterns is still limited. This paper describes an in-process inspection system for the thick film screen printing process where high density complex signal patterns are printed on ceramic substrate. The cost of the process necessitates a high production yield. The quality requirement is stringent, and the alternative, manual inspection, is both inefficient and ineffective. Using existing digitizing tube technology coupled with a mini computer, the system scans and detects defects based on programmable parameters. The computer controlled stage transport allows full coverage of the object under inspection. An algorithm is developed to automatically align the object and the master images. The system represents a successful application of imaging technolopy to industrial inspection of complex and noisy pat-terns. It is fully automated and human engineered such that an operator with a minimum of skills and train-ing can operate a cluster of up to three systems, The high depree of confidence in its inspection efficiency and the throughput rate render the system a viable industrial inspection tool.© (1978) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Binary morphological processing operations have been shown to be useful in a range of industrial inspection applications. These operations have been available in two forms: As software-based operations on general-purpose algorithm development workstations, such as Bruce Batchelor's SUSIE and its many descendants, and as hardware-based operations on dedicated "morphology engines." Obstacles to wider use of these methods include • Limited familiarity with binary morphological processing techniques on the part of general users, • The slow operation of software-based implementations, and • The relatively high initial cost of fast special-purpose binary morphological processing hardware. This paper describes another alternative: A near-real-time implementation of binary morphological processing as part of a very large set of operations on a moderately-priced general-purpose image processing and algorithm development workstation. The workstation is based on commercially-available image processing boards, and provides a high-level operator interface. With this system, the speed depends on the size of the structuring element, but migration to real-time implementations using the same hardware family is straightforward. This implementation allows arbitrary specification of the final structuring element, with no constraints as to symmetry, connectedness, concavities, holes, etc. This eliminates one of the obstacles to the use of complex structuring elements on some dedicated morphological processing machines -- the need to find a set of simple "primitives" which, when applied sequentially, will yield the desired structuring element. The nature of the implementation also allows for a new possibility: By a change in one register, operation is converted from a strict binary morphology into "fuzzy" morphology (not grey-level morphology), in which the user can specify a "percentage fit" of structuring element to image. By this means, the structuring element can still be made to "fit" even in the presence of some "noise" pixels, without the need for separate noise-removal steps. This greatly increases the robustness and practicality of morphological methods in "real world" industrial inspection applications.

Traditional tube-based x-ray sources are widely employed in medical imaging, security screening, and industrial inspection. The cone-beam produced by these tubes is simple to apply, but often demands a long stand-off distance to the object of interest. When combined with the bulk of tubes and their attendant power supplies and cooling systems, the footprint requirement of traditional sources often impedes their use, especially in mobile situations. Here we present an approach to a distributed, flat-panel x-ray source, which eliminates the aforementioned bulk, weight and need for standoff. This source uses spontaneous polarization in pyroelectric crystals to generate high fields and field enhanced emission from micropatterned tips to create a large array of electron beamlets. When combined with a transmission Bremsstrahlung target, a mechanism for raster control of the emitters, and a collimator, this source offers a new and cost effective way to perform stationary and portable imaging. The working principles and performance characteristics of this source are presented. The demands placed on the imaging detector and image processing are also described. Finally, prospects for new promising applications (such as mammography) are mentioned.

Texture analysis has become an important task in image processing because it is used as a preprocessing stage in different research areas including medical image analysis, industrial inspection, segmentation of remote sensed imaginary, multimedia indexing and retrieval. In order to extract visual texture features a texture image analysis technique is presented based on the Hermite transform. Psychovisual evidence suggests that the Gaussian derivatives fit the receptive field profiles of mammalian visual systems. The Hermite transform describes locally basic texture features in terms of Gaussian derivatives. Multiresolution combined with several analysis orders provides detection of patterns that characterizes every texture class. The analysis of the local maximum energy direction and steering of the transformation coefficients increase the method robustness against the texture orientation. This method presents an advantage over classical filter bank design because in the latter a fixed number of orientations for the analysis has to be selected. During the training stage, a subset of the Hermite analysis filters is chosen in order to improve the inter-class separability, reduce dimensionality of the feature vectors and computational cost during the classification stage. We exhaustively evaluated the correct classification rate of real randomly selected training and testing texture subsets using several kinds of common used texture features. A comparison between different distance measurements is also presented. Results of the unsupervised real texture segmentation using this approach and comparison with previous approaches showed the benefits of our proposal.

The quality of optics' surface is of an important character which has critical influence in optical system. There are many ways to detect surface quality, but interferometry is considered as an effective technology. Interfere fringes which obtained from our experiment based on equal thickness interference were processed with digital image processing (DIP) technology in this paper. Image smoothing, fringes thinning, fringes' space measurement and P-V value measurement were done in this processing, Which can give optics' surface information automatically and accurately. We measured some optics whose apertures are not more than 30mm, and the result is satisfactory. Our method will be widely used in industrial inspection, especially in optical works in the future.

O-rings are regularly replaced in aircraft and if they are not replaced or if they are installed improperly, they can result in catastrophic failure of the aircraft. It is critical that the o-rings be packaged correctly to avoid mistakes made by technicians during routine maintenance. For this reason, fines may be imposed on the o-ring manufacturer if the o-rings are packaged incorrectly. That is, a single o-ring must be packaged and labeled properly. No o-rings or more than one o-ring per package is not acceptable. We present an industrial inspection system based on real-time composite correlation filtering that has successfully solved this problem in spite of opaque paper o-ring packages. We present the system design including the composite filter design.

The visual inspection technology, which can extraordinarily reduce inspection cost, greatly speed up of development and raise the efficiency of production, has become the topic of the day in industrial online inspection in recent years which is widely used in the field of industrial inspection 1 . In this paper, a method of high accuracy and long size base on machine vision is developed, and some key techniques are introduced in detail, such as lighting system, optical lens, camera , and image processing algorithm. The quality of image is improved by optical imaging and lighting system. A CMOS (Complementary Metal-Oxide Semiconductor) camera with a selected optical lens was used to capture the image of an object. The analogue signals of the image were transformed into corresponding digital grey level values. By using the binarization technique, the images of background and the object were segmented. The precision of edge detection was improved by sub-pixel technique. The precision of fitting was improved by least square method and Hough transformation. Space coordinate was transformed from an image pixel coordinate through coordination. An automatic detection system of ceramic pieces was also developed .And by taking example of ceramic piece 80 mm long and 80 mm wide, the test method is applied by the detection system with accuracy about 8um.The outcome of experiment accords with the fact, this method is simple and feasible, and is of high accuracy, dependability, and high automation.

Pattern matching has long been a cornerstone of industrial inspection. For example, in order to obtain high accuracy, modern overlay metrology tool optics are optimized to ensure symmetry around the central axis. To obtain best performance, the metrology target should be as close as possible to that axis, hence a pattern recognition stage is usually used to verify target position before measurement. However most of the work performed to date has concentrated on situations where the imaging process could be described by simple ray-tracing, where the image is formed by albedo difference between surfaces rather than interference. However, current semiconductor technology requires optical identification of targets less than 30 microns (i.e. about 50 wavelengths) across, and of order 1 wavelength deep, and this description is no longer valid; interference and focusing effects become dominant. In this paper we examine these effects, and their impact on a number of different techniques. We compare image-based and CAD-derived models in the training of the pattern recognition system; CAD-derived models are of particular interest due to their use in “imageless” recipe creation techniques. Our chief metrics are precision and reliability. We show that for both types of pattern matching approach, submicron precision and high reliability is achievable even in very challenging optical environments. We show that, while generally inferior to image based models, that models derived from design data are more robust to changes caused by process variation, namely changes in illumination, contrast and focus.

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.

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.

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 comparison between the features obtained using Gabor filtering on complex images under illumination and focus variations.

Fast and robust 3D quality control as well as fast deformation measurement is of particular importance for industrial inspection. Additionally a direct response about measured properties is desired. Therefore, robust optical techniques are needed which use as few images as possible for measurement and visualize results in an efficient way. One promising technique for this aim is the inverse pattern projection which has the following advantages: The technique codes the information of a preceding measurement into the projected inverse pattern. Thus, it is possible to do differential measurements using only one camera frame for each state. Additionally, the results are optimized straight fringes for sampling which are independent of the object curvature. The ability to use any image for inverse projection enables the use for augmented reality, i.e. any properties can be visualized directly on the object's surface which makes inspections easier than with use of a separated indicating device. The hardware needs are low as just a programmable projector and a standard camera are necessary. The basic idea of inverse pattern projection, necessary algorithms ane found optimizations are demonstrated, roughly. Evaluation techniques were found to preserve a high quality phase measurement under imperfect conditions. The different application fields can be sorted out by the type of pattern used for inverse projection. We select two main topics for presentation. One is the incremental (one image per state) deformation measurement which is a promising technique for high speed deformation measurements. A video series of a wavering flag with projected inverse pattern was evaluated to show the complete deformation series. The other application is the optical feature marking (augmented reality) that allows to map any measured result directly onto the object under investigation. The general ability to straighten any kind of information on 3D surfaces is shown while preserving an exact mapping of camera image and object parts. In many cases this supersedes an additional monitor to view results and allows an operator to investigate results on the object, directly.

In industrial inspection one of the key areas is detection of defects from textured surfaces. The goal is to differentiate between a good, normal surface texture and a defected surface texture. In this paper this is achieved with a two-class classifier that is taught only with fault-free samples of surface texture. An unsupervised segmentation scheme is formulated where an unknown sample is classified as a defect if it differs enough from the estimated distribution of texture features extracted from fault-free samples. The extension of the self-organizing map (SOM) algorithm, the so-called statistical SOM, is used to estimate the distribution. Different versions of the statistical SOM are introduced and their computational requirements are discussed. The proposed methods are shown to perform well in segmentation of texture surface images with different kinds of defects.

The traditional image processing techniques require a lot of computational effort due to data on each pixel are computed in a sequential way and the path of information is an A/D converter. The delay accumulation create in this process is unacceptable in real time image processing because of the high information flow managed in the usual vision tasks (e.g. automatic industrial inspection, vision problems in robotics, pattern analysis, etc. ). Thus, the use of a massive parallel architecture working with analog signals avoids the previous problems. This is just the basis idea of Cellular Neural Network (CNN’s): an array of analogic dynamic processors which cells interact directly within a finite local neighborhood . The local CNN connectivity allow its realization as VLSI chips that can operate at a very high speed and complexity. Nowadays CNN architectures implemented as VLSI chips shows the aptitude of extremely high speed compared with traditional digital image processing tools. The proliferation of more and more sophisticated CNN architectures, and the increasing effort to implant practical system based in CNN chips, make important the development of analog algorithm to perform complex image processing tasks dedicated to many different fields, i.e. industrial applications, robotic systems and pattern recognition. The objective of this work is to generate a learning machine capable of find solutions for complex image processing task by CNN’s. First a general machine for automatic analog algorithm design independent of the problem to solve is created, this is accomplished through an evolutionary strategy that is an extension of genetic programming. Second, this work introduces a suite of sub-mechanisms that increase the power of genetic programming and contribute to reduce the enormous space search for producing a plentiful search. Some concepts in this section are related with AI theory, in such a way that in this work we are in the intersection field of AI and Image Processing by CNN. Keywords: Cellular Neural Network, Visual Processing, Parallel Computation, Real Time Processing, Automatic Templates Generation, Genetic Algorithms.

In car assembly plants, when the car assembly process has finished, it is necessary to carry out a registration process which certifies that all the identification labels placed in specific parts of the vehicle (motor, chassis, board, windshield, right front post, etc.) correspond with a particular car unit. In this paper we propose a system for capturing and processing car identification labels. The system captures the required images for the registration process using a CCD camera with an appropriate mount. Images are analyzed and stored by the system, generating a report, where all the captured visual information is included. The information is further stored in a way, that allows the retrieval of each report. Results and discussion about the system are presented. Keywords: Machine Vision, Industrial Inspection, Car Registration 1. INTRODUCCTION In car assembly plants, when the car assembly process has finished, it is necessary to carry out a registration process which certifies that all the identification labels or VIN (Vehicle Identification Number) placed in specific parts of the vehicle (motor, chassis, board, windshield, right front post, etc.) correspond with a particular car unit. VIN numbers are used to identify uniquely vehicles around the world. They codify information about country of origin, manufacturer, assembly factory, and specific characteristics of a particular vehicle. The registration process is useful for example, when a car accident arise, and the factory or the driver wants to know if a determined part of the car has passed the quality tests without errors. Other situation when the registration process is useful , occurs when a car is stolen and the motor number is erased and minted again. A system can discover this change if it checks the other identification labels for consistency among themselves and with the car characteristics (model, color, etc.). 1.1. The identification labels description

Analog CNN array computer arises as an alternative to traditional digital processors in many industrial inspection like visual quality control of metal laminants, capable of make in a single chip Tera equivalent operations per second. A 4096 analog CNN processor array is able to perform complex space-time image analysis, being much faster than a camera- computer system in continuous inspection applications. Both chips have been implemented in CMOS technology and they are managed by a 32-bit high-performance low-cost micro- controller that closes the pan, tilt, lighting, focus and zoom loops required in the implementation of the active vision strategies. Several convolution masks for the Cellular Processors has been selected to detect particular changes in the texture, size, direction or orientation of the image entities, reprogramming `on the fly' the pixel resolution of shape when necessary. Laboratory results present these Cellular Processors and multiple resolution imager circuits as a promising architecture for visual inspection of industrial processes in real time.

The purpose of image analysis as a special type of signal analysis is to obtain relevant information from the physical world for decision making, like in industrial inspection, verification, alarm generation, for taking actions, like control of production systems, vehicle control, or for data transformation, like 3D-reconstruction, image coding, image rectification, etc. Any type of image analysis involves data grouping and data interpretation. The classical approach is the bottom-up strategy which is very efficient but requires extra efforts to obtain a high image quality and is quite sensitive even to slight changes in the scene setup, noise, disturbances, illumination effects, etc. The current trend in image analysis is to cope with more complex situations, like time varying scenes, spatial scenes, where the image depends on the viewing direction, or cases where the quality in one single image is too low to permit the extraction of reliable information. In these situations image analysis requires support from other sensors at different viewing angles, recordings form different time instances or the exploitation of different physical principles for image generation, i.e. multisensor data and the systematic use of knowledge. More flexibility in the analysis method is obtained by combining the classical data driven bottom-up strategy with the expectation driven top-down strategy. These new approaches will be illustrated in the paper by three examples, the analysis-by-synthesis strategy, multisensory analysis and knowledge based analysis.

This paper presents a method for minimizing the total elapsed time spent by n tasks running on m differents processors working in parallel. The developed algorithm not only minimizes the total elapsed time but also reduces the idle time and waiting time of in-process tasks. This condition is very important in some applications of computer vision in which the time to finish the total process is particularly critical -- quality control in industrial inspection, real- time computer vision, guided robots. The scheduling algorithm is based on the use of two matrices, obtained from the precedence relationships between tasks, and the data obtained from the two matrices. The developed scheduling algorithm has been tested in one application of quality control using computer vision. The results obtained have been satisfactory in the application of different image processing algorithms.

communications is demonstrated in this paper. The processing is done in the space domain utilising zero phase digitalfilters. In this paper, a recursive pole-zero placement digital filter design method has been employed, and the filters havebeen applied to both 1-D simulated signals and real 2-D images. The results are explained and suggestions to furtherimprove either the quality or speed of the method are also presented. Two methods of producing zero-phase digital filtersare explained which is important in preserving the integrity of an image. Theoretical derivation and detailed analysis of thedemodulation method and filter design are presented. Images are included as to give a better understanding of the approach.Finally, comments, improvements and comparisons of the method are included.Keywords : digital filters, phase demodulation, shape measurement, pole-zero placement.2. INTRODUCTIONNon-contact 3-D phase measurement techniques are now in a mature stage of development and have been widely usedin both high precision measurement and industrial inspection. The requirements or criteria of the techniques used in thesetwo environments are similar but with subtle differences. Various approaches or methods have been proposed and developedin the area including Fourier Transforms1, Temporal and Spatial Phase-Stepping2'3, and Shadow Moire4. Among thetechniques, Phase-Stepping or Phase-Shifting are the most computational inexpensive and simple approach but they requirethree to five images to be taken for the processing and so, they are not suitable for use in the adverse environments usuallyfound in industrial sites, but are good for high precision measurements. The Fourier Transform method needs only one

Binary correlation is often used for finding specified patterns in complex binary images, especially in industrial inspection tasks such as locating the corners and/or edges of parts. As such, it is an important tool for higher-level 'intelligent' vision systems. Binary correlation is a form of binary template matching which provides a numerical value corresponding to 'degree of fit' rather than an 'all or nothing' answer. Commercially available high-speed image processing systems can readily perform this operation using linear convolvers, but such convolvers are very expensive except for very small kernels. Furthermore, linear convolvers constitute a gross 'overkill' for the relatively simple operation of binary correlation. Specialized binary convolvers have been built, but are not part of standard commercial systems. This paper describes a new pipelined implementation of binary correlation which fits into the standard SKIPSM (separated-kernel image processing using finite state machines) architecture and which can be built using standard ICs costing less than $500 total. The same approach can also be implemented in software, providing an order-of-magnitude increase in speed at no extra cost. Furthermore, this same SKIPSM architecture is highly versatile and programmable, allowing it to be software-reconfigured to perform hundreds of other pipelined image processing operations.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Object segmentation, recognition and localization are challenging because of the large amount of input data and because of the invariances required. We discuss strategies to overcome these problems, considering sensors, algorithms and architectures. Specifically, we address neural nets and Hough strategies. The ability of implicit learning makes neural nets interesting for industrial inspection: compared to classical methods they promise robustness against variations of the input data. Furthermore, no expert is necessary for supervision. The inherent parallelity simplifies the design of algorithms. However, the advantages are counterbalanced by a serious drawback: the high computational complexity -- if images are considered. The ability of optics, to help by its inherent parallelity is limited, because neural architectures are usually space variant and cannot simply be implemented optically. We discuss approaching these problems by feature extraction, by sparse algorithms and by space invariant architectures. A competitive strategy for object recognition and localization is based on probability tables, such as the Hough transform uses: a couple of weak but independent hypotheses can give a safe decision about the kind and the locus of an object. This method requires a learning phase prior to the working phase, as the neural strategy does. In that sense it is similar, however, the computational complexity can be much smaller. This makes it possible to segment, localize and recognize objects invariant against shift, rotation and scale.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This work concerns computationally-efficient computer vision methods for the search for and identification of small objects in large images. The approach combines neural network pattern recognition with pyramid-based coarse-to-finesearch, in a way that eliminates the drawbacks of each method when used by itself, and in addition, improves object iden-tification through learning and exploiting the low-resolution image context associated with the objects. The presentationwill describe the system architecture and the performance on illustrative problems. 1. INTRODUCTION This paper concerns computationally-efficient computer vision methods for ATh applications, especially those in whichthere are multiple kinds of relatively small targets sparsely distributed throughout large images (see Fig. 1). The portion ofthe image occupied by the targets is frequently quite small. Consequently, the great majority of the image data represents"clutter", i.e., irrelevant variations that must be ignored. Applications of this type include reconnaissance, surveillance,target acquisition and tracking, robotics, collision avoidance and industrial inspection. Our approach to these problemshas three integrated components: coarse-to-fine search through multi-resolution image pyramids; neural networks for pat-tern classification; and fusion of low-resolution context information with high-resolution pattern data.Figure 1. Example of an ATR scene and target. The small pattern (upper right) is an image of a target,to be found in the scene shown below it.

A robotic vision system design for industrial inspection is proposed. It implements a 2D color CCD camera with a modified focusing lens. A diffractive optic element (DOE) is etched into a small, off axis, region of the lens. A similar region of the detector is dedicated to the DOE image. This image, after some processing provides the third dimension. The output information includes a standard 2D image, the distance to a component under test and if the object is in motion, the direction and velocity towards or away from the detector plane. This system can be designed to operate at visible or infrared wavelengths for a variety of inspection applications.

(Xi) -filters were introduced in the mid-1980s by the author and have been used in many applications, including industrial inspection, remote sensing, and medicine. This paper reviews some new developments in each of these areas. The (Xi) -filter is based on high- speed LUT (lookup table) operations wherein small-kernel, nonlinear convolutions are performed on three-dimensional data. In order to execute 3D transforms by LUT, it is necessary to utilize as compact a kernel as possible. This has led to the FCC (face-centered- cubic) tessellation where the kernel comprises only 13 binary data elements or voxels. (This is in contradistinction to the Cartesian tessellation where the kernel comprises 27 voxels.) Since each LUT contains at each of 8192 locations the transformed value of the voxel, the program word that defines a single (Xi) -filter transform is 8192 bits in length. There are, therefore, an essentially infinite number of program words and therefore an infinite number of transforms. In order to delimit the number of transforms to a set that are both useful and manageable, program works limited to the various ranking filters have been generated. In a kernel of 13 elements, there are, of course, only 13 ranks. By iterating (Xi) -filters based on these ranking transforms many interesting operations are possible as illustrated in this paper.

The majority of machine vision systems derive their input data by digitising an image to produce a square grid of sampled points. However, other sampling techniques can represent equal picture information in a smaller number of samples, with a consequent reduction in data rate. Several workers have looked at regular hexagonal sampling of images which produces optimum data rates for a given information content. Previous work on hexagonal sampling by the authors and others, has shown that image processing operators are computationally more efficient, and as accurate, as their square counterparts. Historically, one factor which has lead to the predominance of square sampling in vision systems, is that this produces images which are more visually pleasing to human observers. This paper describes an investigation of machine vision systems performing industrial inspection tasks, which suggests that in such applications, hexagonal systems out-perform square systems. In particular hexagonal operators can follow tight curves more accurately, allowing better surface defect detection. A surprising observation of this work was that with such images, hexagonal sampling also gave images which were more visually pleasing to human operators. The paper presents a study of sampling point geometry and operator design. Details are given of an implementation of a set of hexagonal, grey-scale operators for use in pipeline or other image processing systems, and a comparison of square and hexagonal techniques has been made. Results of operations on real and simulated surface defect images are given for both sampling systems and the requirement for a defect detection figure of merit identified.

Machine vision systems and other automated visual inspection (AVI) systems have been proving their usefulness in factories for more than a decade. In spite of this, the number of installed systems is far below the number that could profitably be employed. In the opinion of the authors, the primary reason for this is the high cost of customizing vision systems to meet applications requirements. A three-part approach to this problem has proven to be useful: 1. A multi-phase paradigm for customer interaction, system specification, system development, and system installation; 2. A powerful and easy-to-use system development environment, including a a flexible laboratory lighting setup, plus software-based tools to assist in the design of image acquisition systems, b. an image processing environment with a very large repertoire of image processing and feature extraction operations and an easy-to-use command interpreter having macro capabilities, and c. an image analysis environment with high-level constructs, a flexible and powerful syntax, and a "seamless" interface to the image processing level; and 3. A moderately-priced high-speed "target" system fully compatible with the development environment, so that algorithms developed thereon can be transferred directly to the factory environment without further development costs or reprogramming. Items 1 and 2 are covered in other papers1,23,4,5 and are touched on here only briefly. Item 3 is the main subject of this paper. Our major motivation in presenting this paper is to offer suggestions to vendors developing commercial boards and systems, in hopes that the special needs of industrial inspection can be met.

The Fourier Transform Method (FTM) of fringe pattern analysis is one of the most important phase measuring techniques. Its chief advantage is that it requires only one interferogram which is crucial for industrial inspection. The goal of a fully automated phase retrieving technique, however, remains still elusive. Several refinements in 2-D FTM are proposed to achieve fully automatic, high-accuracy analysis of a fringe pattern. The new possibilities of on-line analysis ensuing from the recent advances in microcomputer technology are discussed.

A set of logical and algebraic image operations is introduced. These image operators include operators for binary, silhouette, and grey-scale imagery. Simple illustrations of these image operators are given. Examples of industrial inspection and robotic visual feedback problems are considered. Results of object identification and defect locations obtained using these image operations are presented.© (1982) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.