Some of the ethical, environmental and social issues relating to the design and use of machine vision systems in manufacturing industry are highlighted. The authors' aim is to emphasize some of the more important issues, and raise general awareness of the need to consider the potential advantages and hazards of machine vision technology. However, in a short article like this, it is impossible to cover the subject comprehensively. This paper should therefore be seen as a discussion document, which it is hoped will provoke more detailed consideration of these very important issues. It follows from an article presented at last year's workshop. Five major topics are discussed: (1) The impact of machine vision systems on the environment; (2) The implications of machine vision for product and factory safety, the health and well-being of employees; (3) The importance of intellectual integrity in a field requiring a careful balance of advanced ideas and technologies; (4) Commercial and managerial integrity; and (5) The impact of machine visions technology on employment prospects, particularly for people with low skill levels.

The World Wide Web Initiative has provided a means for providing hypertext and multimedia based information across the whole Internet. Many applications have been developed on such http servers. One important and novel development on the World Wide Web (WWW) has been the development of computer vision and image processing related courseware facilities and indeed image processing packages. This ranges from the provision of on-line lecture notes, exercises and their solutions to more interactive packages suited primarily for teaching and demonstration packages. Within the WWW there are many pointers that highlight more research based activities. This paper addresses the issues of the implementation of the computer vision and image processing packages, the advantages gained from using a hypertext based system, and also relates the practical experiences of using the packages in a class environment. The paper addresses issues of how best to provide information in such a hypertext based system and how interactive image processing packages can be developed. A suite of multimedia based tools have been developed to facilitate such systems and these are described in the paper. A brief survey of related sources of information on the World Wide Web also is presented.

Machine vision systems are widely employed for inspection of blister-packaged pharmaceutical products, and the quality of inspection which can be performed by any vision system is directly related to the quality of images obtained. This paper covers design and selection of illumination, optics, and camera technology to optimize image quality for inspection of blister-packaged tablets and capsules. The nature of image quality is discussed in terms of the optical properties of the materials involved and the machine vision inspection techniques to be applied, and practical application guidelines are presented.

There is an increasing acceptance of reduced waste packaging methods for pharmaceutical and medical products. The high level of product integrity must be maintained, while manufacturing lines are required to increase production rates. To ensure their confidence in these packaging methods, manufacturers have turned to process validation as one method of check. In addition to that effort, automated on-line inspection has become increasingly important. Automated inspection can be used to augment manual inspection techniques that are viable at slower production rates. In this paper we explore the elements of a systematic approach that can provide 100% automatic inspection of product seals at full production rates. The various materials used to seal packages effect the system configuration. One such package sealing material is highly specular (mirror-like) laminated foil. A characteristic of this packaging method is its ability to reflect nearly all of the light from the surface. However, the heat process required to bond the seal to the package creates a coining effect where a uniform, low to medium intensity light source, transmitted at a low incident angle, can be used to identify seal defects. It is equally difficult to inspect package seals that are opaque, translucent, or transparent. Each seal material requires a specific lighting solution. When using reflective material, great care must be taken to develop and integrate the lighting method to an automated package seal inspection system.

Pharmaceutical manufacturers now rely on machine vision for effective automated inspection of critical components in the packaged product. Several years of trial and error, followed by FDA influence, aided in designing a procedure for successful implementation. Difficult, high speed applications require unique technology and clear project planning.

This paper describes digital image processing techniques that can be used for on-line quality assurance in web material manufacturing. The primary concentration of this work is on describing properties of the random texture patterns formed by melt blowing polypropylene resin. The proposed descriptors characterize statistical properties of a texture pattern and are related to material performance. It is envisioned that the descriptors will be used eventually to establish quantitative relationships between the microscopic properties of the material and its performance.

Performing complex image processing operations on digital signal processing (DSP) chips offers a great deal of flexibility compared to hardware based solutions, although the later are generally much faster. In this paper the authors fully evaluate the TMS320C40 chip which is a state of the art DSP chip specially designed for highly intensive computations such as image processing, with the extra flexibility of having in-built communication ports to enable parallel processing.

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.

Skeletonization of binary images is an essential step in the inspection of many products, most notably printed circuit boards. It also is used in many other situations, an unusual example being the location of branching points on growing plants for purposes of cutting and vegetative propagation. Commercially available image processing boards typically cannot perform this operation, although they readily perform the easier task of repeated binary erosion. It has been shown that, in addition to hundreds of other image processing operations, the inexpensive standard SKIPSM (separated-kernel image processing using finite state machines) architecture can be used to implement binary skeletonization in one pipelined pass per stage of erosion. This paper considers the feasibility of extending the SKIPSM skeletonization implementation to obtain 'hairless' skeletons by using Jang and Chin's algorithm instead of Floeder's algorithm as used in the first implementation.

This paper presents a typical industrial application of machine vision in order to classify and select different types of cylindrical steel bars in combination with direct process control. The main classification algorithm consists of a combination of several routines, using different image processing methods. On the one hand a textural approach, using first and second order statistics, is used. Typical histogram data in addition with gray level dependence matrices give some textural classification criteria. On the other hand the search and utilization of geometric criteria supply additional features for classification. Several contour measurement routines deliver a set of additional information about the examined bar. The paper offers details about the used classification algorithms. Furthermore it deals with experimental results such as velocity and rate of selection success.

A new method of implementing a wide range of standard image processing operations in a real-time finite-state-machine architecture has been presented at various conferences in the past year. This architecture, under the generic name SKIPSM (separated-kernel image processing using finite state machines), has been shown to be capable of carrying out binary morphology with very large arbitrary structuring elements, simultaneous application of many binary structuring elements, gray-level morphology, binary and gray-level template matching, binary skeletonization, binary correlation, row and column summations, and many other operations. This paper describes inexpensive hardware implementations of the SKIPSM architecture, including a daughter board compatible with commercially available pipelined image processing hardware.

The paper presents a hybrid approach to the problem of classifier construction for machine vision based inspection systems. The method allows the user to integrate different types of classifiers and exploit different sources of information such as sample data and expert knowledge. Of particular interest is the problem of classification reliability in the case of small training sets. A novel forward fuzzy decision tree induction method is proposed to handle different types of uncertainty. The performance of the method is compared experimentally with other classifiers using artificial and machine vision data.

The advent of low-cost charge coupled device (CCD) cameras offers an alternative approach to the construction of visual inspection systems in the manufacturing industry. An array of such cameras may be arranged such that a panoramic view may be achieved without the use of any mechanical components such as motors, thereby reducing the need for maintenance. Also, as the arrangement is completely electronic, such a system lends itself directly to advanced automation. This paper sets out to describe the components necessary in such a system.

Contrast based texture segmentation method is presented in this paper. When defining the measurements such as contrast and homogeneity of regions the visual characteristics are considered. When calculating the measurements some fast algorithms are used. Some images and their segmented results are also provided.

PCI bus mastering with a frame grabber provides a 40 times speed improvement over ISA bus frame grabbing while lowering frame grabber costs by more than 50%.

The paper presets the redundant Hough transform (RHT). It differs from classical Hough transform (HT) by the 'dual voting' in the HT-plane. Mapping of each straight-line segment from image-plane to two points in HT-plane is produced. The using of RHT allows us to reduce translation and turn distortions in image-plane to cyclic shifts along coordinate axes in HT-plane for all values of the distortions. On the basis of offered RHT, algorithms for machine vision systems can be constructed which provide invariant to turn and translation distortions recognition with estimation of that distortion's values. The block diagram of a specialized processor for RHT realization is offered.

The stability of glass substrates is an important concern for the flat panel display industry. High-resolution displays have very tight geometrical requirements and alignment of the various display components is critical if good performance is to be obtained. Prior to development of manufacturing processes for these displays, it is necessary to determine how glass substrates change during the various processing steps. This paper describes a system to measure electrode patterns before and after critical processing steps for color plasma panels. The electrode patterns, which are made of thin-film gold, are a series of parallel electrodes. In order to measure electrode locations, a vision system consisting of an X-Y stage, a video camera, a frame grabber, and PC-compatible computer was used. Images captured with this setup were processed to minimize the effects of noise and improve accuracy. A gray-scale interpolation technique in which the centroids of the electrodes are calculated was used to enhance measurement resolution.

This paper discusses a high-speed 3-D inspection system for solder-bumps. The system uses a high-speed 3-D sensor system and an accurate measurement algorithm. Solder-bumps have recently been used for flip-chip bonding. Before bonding all bumps need their height and diameter inspected and if bumps are too big or too small, there is a danger of short or open circuits occurring after bonding on the substrate electrodes. Thus, a 100% inspection is required to assure high flip-chip bonding process yields. We developed a laser-based high- speed bump height capture system and an accurate bump height and diameter measurement algorithm. The inspection system takes 20 milliseconds to measure the height and diameter of a bump. It measures the bump height to an accuracy of plus or minus 3 micrometer, and the bump diameter to plus or minus 5 micrometer. Thus, this system is suitable for performing a 100% inspection of solder-bumps.

Developers of machine vision systems for industrial applications are frequently exposed to the problem of proving to their customers that specified performance measures are met. A typical example would be the rate of correct classification in defect detection machines that usually will be in the range of 95 - 100%. We call such machines near perfect. In practice this figure is stated for the complete inspection decision, which in general is based on a number of subdecisions made by the machine. An example would be surface inspection in industrial production, where a workpiece will be rejected if one or several defects are detected. Let's assume that the probability of false classification for a single defect is p₁. In the case where several defects may appear on the surface every defect contributes to the final decision, with the probability of a wrong decision p₂. It would appear logical that p₂ is larger than p₁, because the more defects are found on the surface, the more likely the system would make a wrong decision (all the p₁s for the single defects would add up). In this paper we show that although seeming paradox the reverse is true. We show that with estimates of p₁, the joint decision can be optimized such that the actual error rate of the defect detection machine is less than p₁. We also give practical recommendations on how to tune the pattern recognizers to achieve optimal performance.

As the recently proposed JTC has been proven to be effective for implementation of a real- time target tracking system, the interest in the electronic support system for the real-time JTC tracker has been increased. Accordingly, we propose a tracking system which is based on BPEJTC and adaptive to the fixed site. But because the EOTS is generally needed in the moving site such as with aircraft and vehicles, and there are many different tracking algorithms to adopt the BPEJTC, we present an advanced version of BPEJTC driver which has synchronization input so as to be used for the target pointer. In addition to the designed system architecture, some experimental results conducted by this system are illustrated.

A new method of automatic measurement of a pinion type gear's profile using computer vision is provided and the measurement principle is analyzed. Based on a tool maker's microscope, a CCD camera, a real-time image sampling board, and a computer a vision system is established. Through calibration, edge detection, and detection of pinion gear center two kinds of modes, i.e., centering measurement and non-centering measurement of profile error, are realized by special software. A new operator is provided to find the edge of the binary digital image and the results of edge detection by the operator are presented. Block gauge method is used in system calibration and the results show that the system resolution will reach 3.28 micrometer/pixel (horizontal) and 1.60 micrometer/pixel (vertical) if the 5 (Chi) objective is used. The measurement results of an involute pinion type gear's profile error is also given in the paper.

An inexpensive but versatile human-computer interface (HCI) for a machine vision system is described. Widely available hardware and computing components are controlled by software based on HyperCard and Prolog. While considerable benefit is obtained using just one of these programming tools, it has been found that the combination provides many advantages, including ease of use and great flexibility. Details of what is possible using HyperCard and Prolog individually and both working in harmony are discussed.

PIP (Prolog image processing) is a system currently under development at UWCC, designed to support interactive image processing using the PROLOG programming language. In this paper we discuss Prolog-based image processing paradigms and present a meta-interpreter developed by the first author, designed to support an approach to image processing in PIP which is more in the spirit of Prolog than was previously possible. This meta-interpreter allows backtracking over image processing operations in a manner transparent to the programmer. Currently, for space-efficiency, the programmer needs to indicate over which operations the system may backtrack in a program; however, a number of extensions to the present work, including a more intelligent approach intended to obviate this need, are mentioned at the end of this paper, which the present meta-interpreter will provide a basis for investigating in the future.

This is the latest in a series of publications which develop the theme of programming a machine vision system using the artificial intelligence language Prolog. The article states the long-term objective of the research program of which this work forms part. Many but not yet all of the goals laid out in this plan have already been achieved in an integrated system, which uses a multi-layer control hierarchy. The purpose of the present paper is to demonstrate that a system based upon a Prolog controller is capable of making complex decisions and operating a standard robot. The authors chose, as a vehicle for this exercise, the task of playing dominoes against a human opponent. This game was selected for this demonstration since it models a range of industrial assembly tasks, where parts are to be mated together. (For example, a 'daisy chain' of electronic equipment and the interconnecting cables/adapters may be likened to a chain of dominoes.)

The Gnu project has provided a substantial quantity of free high-quality software tools for UNIX-based machines including the Gnu C compiler which is used on a wide variety of hardware systems including IBM PC-compatible machines using 80386 or newer (32-bit) processors. While this compiler was developed for UNIX applications, it has been successfully ported to DOS and offers substantial benefits over traditional DOS-based 16-bit compilers for machine vision applications. One of the most significant advantages with Gnu C is the removal of the 640 K limit since addressing is performed with 32-bit pointers. Hence, all physical memory can be used directly to store and retrieve images, lookup tables, databases, etc. Execution speed is generally faster also since 32-bit code usually executes faster and there are no far pointers. Protected-mode operation provides other benefits since errant pointers often cause segmentation errors and the source of such errors can be readily identified using special tools provided with the compiler. Examples of vision applications using Gnu C include automatic hand-written address block recognition, counting of shattered-glass particles, and dimensional analysis.

This paper describes the implementation of a high speed inspection system for the wood processing industry. This was realized by a major European manufacturer of wood processing equipment, who is a customer of Loughborough Sound Images plc for the DSP based image processing hardware. The system provides quality inspection and control for planks of wood which have been cut from larger logs, and aims to measure the following factors: (1) Knots in the planks caused by branches in the original tree. (2) Cracks in the cut planks. (3) 'Graining' in the cut planks, where the grain pattern is very deep and has affected the surface smoothness of the wood. The range of factors to be measured requires that a combination of novel algorithms be used, and an overview of these is given in the paper. A major part of the implementation is the overall system architecture and design, and this also is described. Special consideration is given to how the algorithms were implemented on the equipment chosen and how the considerable technical problems in this particular high speed implementation were solved.

Object-oriented interfaces (OOIs) have become an important component of automation activity. Object-oriented software techniques have provided some hope to cope with the complexity of modern software development tasks. Object orientation is expressed by many researchers as an important direction in designing and implementing software in the 1990s and beyond. In today's electronics industry, there are several different types of interfaces used for different pieces of manufacturing equipment. It is now possible to create a general, OOI for most manufacturing equipment that is easy to use, easy to learn how to use, and easy to modify. Such an interface can benefit the user in terms of savings in time and money. The laser soldering interface, designed and implemented in the Center for Integrated Electronics and Electronics Manufacturing (CIEEM) at Rensselaer, is one of the 'flexible' user interfaces described above. This paper describes the object-oriented graphical tool (OOGT) development and its final structure.

This paper focuses on recent experiences gained in the computer-aided design of image acquisition for demanding surface inspection applications. The design procedure used in the experiments has been developed at VTT Electronics and a commercially available optics design software is used in the practical implementation. An example of the design of image acquisition for metal, paper, and wood inspection is described to illustrate the utility and practicality of the planning tool for assisting in the design of image acquisition. Certain crucial optical parameters, such as the divergence of an illuminator, the diffusion level of a diffuser, and the illumination and viewing angles are considered in particular and their effects on the resulting illumination patterns on the surfaces of objects are simulated with the help of the design tool. Finally, the simulation results are compared with measurements acquired with a real imaging construction and test samples and the advantages and drawbacks of the computer aided planning tool are evaluated.

The artificial retina chip, consisting of an array of variable sensitivity photo-detector cells (VSPD), attains versatile focal plane image processing using vector-matrix multiplication. We designed an n-MOS VSPD with a pn photodiode and a differential amplifier, which realizes programmable positive and negative sensitivity. The pixels of the fabricated 256 multiplied by 256 pixel artificial retina chip are 35 multiplied by 26 micrometers squared with a 25% aperture ratio in a 2 micrometer n-MOS process. The photosensitivity is 0.23 (mu) A/lx for 1 ms accumulation time. We demonstrate image capture in 'video mode' and 'edge extraction mode.'

The necessity of special and system programming strictly limits the possibilities of image processing specialists developing the program applications if they are not simultaneously the professional programmers. In many respects the efficiency of image processing is determined by the user's interface convenience and saturation and its orientation to the decision of a particular applied problem. There is a rather significant problem of the design of a flexible system that can execute specific image processing scheme in the conveyor mode without the human supervising after the preliminary fitting with the use of visual programming. The important property of these systems must be the possibility of the automatic processing changing under the condition of initial data updating. The object-oriented frame approach to image processing is developed in this paper. The essence of the given approach consists in that the image processing scheme represents a semantic network of software frames, such that each of them is an independent object, and over each of them it's possible to execute the separate transformation. All necessary types of software frames are considered and the interaction of objects in such a network provided by means of the message transmissions between frames in accordance with some logic rules is discussed. The way of image processing systems design offered permits the user to use all the advantages of object-oriented programming and window- based user's interface for the image processing. To illustrate the resources of this approach authors worked out the Visual PISoft or Windows image processing system on PC.

A crucial step in the manufacture of vaccines is the verification of their potency. An assay of the potency must be carried out on every batch produced to determine the safety and efficacy of the vaccine. Currently, human inspectors count the number of plaques (holes) in a cell layer in a petri dish to estimate the potency.They must determine whether nearby plaques that have overgrown each other's borders are single or multiple plaques and distinguish between plaques and small tears in the cell layer resulting from the processing operations (the edges of tears differ in appearance from the edges of plaques). Because of the judgments required to make these subtle distinctions, human inspectors are inconsistent. In cooperation with Merck & Co., Inc., the Rutgers University Center for Computer Aids for Industrial Productivity has demonstrated the feasibility of achieving consistent automatic counting of plaques by a prototype intelligent machine vision system. The David Sarnoff Research Center developed materials handling equipment and factory information system interfaces to enable this prototype system to be installed in a quality control facility at Merck. This paper describes the overall operation of the machine vision aspects of the system, including optics, illumination, sensing, preprocessing, feature extraction and shape recognition. Results of initial tests of the system are also reported.

An error concealment scheme for MPEG video networking is presented. Cell loss occurs in the presence of network congestion and buffer overflow. This phenomenon of cell loss transforms into lost image blocks in the decoding process, which can severely degrade the viewing quality. The new method differs from the conventional concealment by its exploitation of spatial and temporal redundancies in large scale. The motion estimation is carried out by registering images within a multiresolution pyramid. The global motion is estimated in the lowest resolution level, and is then used to update and refine the local motion. The local motion is further refined iteratively at higher resolution levels. An affine transform is used to extract translation, scaling and rotation parameters. In many applications where there is significant camera motion (e.g., remote surveillance), the new method performs better than the conventional concealment.

Bread crumb grain was studied to develop a model for pattern recognition of bread baked at Hard Winter Wheat Quality Laboratory (HWWQL), Grain Marketing and Production Research Center (GMPRC). Images of bread slices were acquired with a scanner in a 512 multiplied by 512 format. Subimages in the central part of the slices were evaluated by several features such as mean, determinant, eigen values, shape of a slice and other crumb features. Derived features were used to describe slices and loaves. Neural network programs of MATLAB package were used for data analysis. Learning vector quantization method and multivariate discriminant analysis were applied to bread slices from what of different sources. A training and test sets of different bread crumb texture classes were obtained. The ranking of subimages was well correlated with visual judgement. The performance of different models on slice recognition rate was studied to choose the best model. The recognition of classes created according to human judgement with image features was low. Recognition of arbitrarily created classes, according to porosity patterns, with several feature patterns was approximately 90%. Correlation coefficient was approximately 0.7 between slice shape features and loaf volume.

This paper describes an approach for real time tracking of parametric objects. At the foundation of this problem is the requirement to efficiently solve the hidden surface problem which is essential for most three dimensional applications in robotics and computer vision. The approach presented here incorporates the concept of frame to frame coherence because there is incremental change in the perceived view of projected object boundaries from one instance of motion to another. This directly translates into geometric constraints that can be enforced in parallel.