Non-contact digitization with optical image sensors permits the reproduction of complex free form surface structures with high accuracy in short time. The optical sensor--based on Moire techniques combined with triangulation methods--digitizes up to 300,000 point coordinates within one view. Large objects are typically scanned in steps by controlling the sensor's position with a handling system. The partially overlapping views are connected afterwards in the computer to describe the whole model. The point cloud is exported to CAD/CAM/CAQ systems, which are doing the surface oriented processing. Today's applications range from copy-milling and surface reconstruction up to quality control and assurance. Especially the processing of the large point clouds with several millions of points needs intelligent algorithms in data organization and surface fitting. More automatic processing results in relatively low quality of the reconstructed surfaces whereas highest quality is obtained with experience and more or less interactive manipulation.

This paper describes a flexible 3D sensor developed by British Aerospace as part of a collaborative project, LAIRD (IED 1551), which aims to meet the inspection needs of high technology manufacturing. By combining the principles of computer vision and robotics it has been possible to produce a visual inspection system where both the cameras and object can be automatically orientated to suit specific applications. A method for calibrating such a system is detailed as is a strategy for determining how best to orientate the cameras and the object.

The aim of this system development is to make a study in the field of the machine vision in three dimensions in this two basic aspects, such as the practical side of the industrial applicability and the investigation in work lines with a future projection, since it is proved that most of industrial problems about recognition must be considered from the point of view of the 3D analysis, which is essential to be able to make a right study.

The optimization of re-production processes in the industrial manufacturing has long been the main interest in order to stay competitive. Only lately the idea spreads that optimizing the pro- duction, in the sense of innovation, could bring further improvement. This paper discusses aspects of introducing automation technology into the manufacturing of one-of-a-kind products. The co-ordination of 3D-optical feed-back with a graphic layout-, off-line programming and simulation system, and with a dual robot cell for flame-cutting, arc-welding and assembly is being developed at the BIBA as a demonstrator and testbed. The main emphasis is put on the active structured lighting for optical 3D-measurement with LCD-line projectors. The concept of the coded light approach and the phase shifting method for the 3D- data acquisition are described. With attention to the application in a robot cell for one-of-a- kind production the components of the measurement system are designed as robot based tools. For calibration and inspection purposes operation sequences are planned with the graphic simulation system. For the extraction of special points out of the 3D-clusters a simple and pragmatic method is applied.

One of the important tools for speeding up the prototyping of an new industrial or consumer product is the rapid generation of CAD data from hand-made styling models and moulds. We present a new optical 3D digitizing system which produces in a fully automatic way non- ambiguous, absolute and complete surface coordinate data of very complex objects in a short time. The system named `OptoShape' is based on a projection of sinusoidal fringes with a true grey-level matrix projector. The system measures both non-ambiguous and absolute XYZ surface data with a pronounced robustness towards optical surface properties. By moving the 3D sensor head around the object to be digitized with a 3/5 axes manipulator, multiple range images are obtained and automatically merged into a unified cloud of point coordinates. This set of surface coordinates are transferred to a software package where interactive manipulation, sectioning and semi-automatic generation of CAD surface descriptions are performed. CNC data can also be directly generated from the point surface coordinate data set.

Automatic production of mechanical parts requires quality control, in particular dimensional measurement after carrying out critical production steps. Usually, precision measurements are performed manually or by tactile measuring robots. Imaging sensors may replace tactile sensors, thereby increase speed and offer some special features making dimensional measurements fast enough to follow the production rate. These sensors will only be accepted by the user at the shop floor if they are easy to use. A measuring robot carries out very different operations in comparison with a set-up only used for a single inspection task. Consequently, the applied methods including lighting, imaging, and computing have to be useful for a wide range of measuring tasks. The paper describes the system including a special optical set-up for lighting and imaging being well suited to the needs of dimensional inspection at the shop floor an electronic hardware used for real-time image processing image processing software and related tools.

Quality control and inspection has become a major tool in industrial production. New measurement systems have to be developed to fulfill these control mechanisms without interfering the production process. Because of the necessity of non interfering systems those have to be contactless. These requirements can be fulfilled by optical measurement systems. The following paper will show an example of such an optical measurement system for measuring spatially bent tubes. There are numerous tasks to measure tubes for shape control in the industrial process. Tubes with different bends, curves and straight parts have to be compared with CAD data. These tubes could be flexible and therefore difficult to measure mechanically. The main problem is to fix the tubes. Another requirement is a short measurement time for shape control in the production line. The new tube measurement system is specially designed for 3D measurement of tubes. The features of the tube measurement system are: non contact measurement, no elastic deformation of the tubes by the fixture, fast acquisition tube geometric, flexible fixing unit, short measurement time by using digital image processing, variable measurement volume and high accuracy.

High-Accuracy 3D Image Metrology now plays an increasingly important role in a variety of industrial and non-industrial settings, including in particular the area of industrial quality control. This can be directly attributed to the dramatic improvements that have taken place over the past several years in the techniques and algorithms used, as well as in the hardware that is currently available. This paper first gives an overview of factors influencing the geometric and radiometric characteristics of CCD-cameras and imaging systems used for 3D image metrology, which includes a discussion of various existing problem sources. Second, techniques for the geometric calibration of CCD-cameras for high-accuracy 3D image metrology are addressed. Third, elements of the IMETRICS TP200 System are outlined in order to demonstrate the hardware readily available on the market. Finally, a number of applications are shown in order to show the state-of-the-art. Results from numerous tests demonstrated that accuracies corresponding to 1/50th of a pixel spacing can be attained under severe industrial conditions.

This paper presents some research on the use of dense range data for the automatic inspection of mechanical parts that have free-form surfaces. Given a part to be inspected and a corresponding model of the part, the first step towards inspecting the part is the acquisition of a range image of it. In order to be able to compare the part image and its stored model, it is necessary to align the model with the range image of the part. This process, called registration, finds the rigid transformation that superposes model and data. After the registration, the actual inspection uses the range image to verify if all the features predicted in the model are presented and within tolerance. Free-form surfaces are particularly interesting in that few inspection processes can inspect surface shape across the whole surface. We focus on the inspection of free-form surfaces and present some results concerning the extraction of nominal shape models from dense range data using B-splines and the use of B-spline models of free-form surfaces for the purposes of registration.

Vision systems are being applied to a wide range of inspection problems in manufacturing. In 2D systems, a single video camera captures an image of the object and application of suitable image processing techniques enables information about dimension, shape and the presence of features and flaws to be extracted from the image. This can be used to recognize, inspect and/or measure the part. 3D measurement is also possible with vision systems but requires the use of either two or more cameras, or structured lighting (i.e. stripes or grids) and the processing of such images is necessarily considerably more complex, and therefore slower and more expensive than 3D imaging. Ultrasonic imaging is widely used in medical and NDT applications to give 3D images; in these systems, the ultrasound is propagated into a liquid or a solid. Imaging using air-borne ultrasound is much less advanced, mainly due to the limited availability of suitable sensors. Unique 2D ultrasonic ranging systems using in-house built phased arrays have been developed in Nottingham which enable both the range and bearing of targets to be measured. The ultrasonic/vision system will combine the excellent lateral resolution of a vision system with the straightforward range acquisition of the ultrasonic system. The system is expected to extend the use of vision systems in automation, particularly in the area of automated assembly where it can eliminate the need for expensive jigs and orienting part-feeders.

This paper presents a combined measurement system based on TV Holography for vibration measurement and structured light for surface form measurement. A common CCD camera is used for both measurements. In both cases phase stepping techniques have been developed for automation of the analysis. The vibration measurements have been achieved using an 8 frame stereoscopic algorithm which can analyze complex vibrational modes. The two data sets are combined in a wireframe representation and animated to provide a direct visual interpretation of the information. It is necessary to measure complex vibration modeshapes and the surface form of a structure in order to predict noise radiation. Noise radiation predictions are of increasing importance in the design and optimization of motor vehicle structures to achieve desired levels of refinement. Noise radiation can be measured directly using the appropriate microphones and analysis equipment. However, such techniques provide no direct information on the source of the noise, or provide an insight for an approach to reduce the noise level.

The passive stereo vision systems is one of the most general techniques of obtaining 3D information from objects. A common problem in stereo systems is finding the corresponding points in the two images of a scene. To cope with this problem, edge-based stereo methods have been considered useful. For curved surfaces, however, edge-based stereo cannot be applied because of the difficulty of extracting edges from curved surfaces. In this paper, we propose a new stereo algorithm based on isodensity lines extracted from a curved surface. Isodensity lines are extracted by tracing pixels having the same density. The proposed algorithm consists of two processes. The first is the process of finding corresponding isodensity lines between two images. The second process is finding one-to-one pixel correspondence along corresponding isodensity lines, for fine matching. With this algorithm, objects with curved surfaces can be reconstructed--even objects having no edges--and poorly corresponding points can be corrected easily. Though some problems are still to be investigated thoroughly, computer simulation results show good accuracy. From the experimental results, the use of this method for the 3D reconstruction of objects with curved surfaces can be considered a serious possibility.

As machine vision systems become more complex they may use several different types of cameras within one system. This paper discusses a recent development at DALSA to provide the three main sensor technologies (line scan, area scan, TDI scan) within one camera family by using a modular approach to the design. The camera family provides the same mechanical and electrical interface for all three sensor types. This reduces the complexity of integrating various camera types into one system. The new camera uses a custom designed application specific integrated circuit to perform the various logical control required by the three sensor technologies. The camera system has a compatible internal bus for simple communication interface between modular functions. An example of one module which will be discussed is an internal digitizing circuit which provides 8 bits of digital video at 16 MHz within the basic camera unit.

Following a need of change from large to small dimensions in technology, research and manufacturing, new light-weight optical tools are required. Based on recent advances in chip- technology and microtechnique, small CCD-cameras, able to be integrated in handy optical instruments, were developed. The fiber optical hand microscope, supplied by a remote light source with incorporated video-control unit, capable of automatic or manual light flux adjustment, guarantees constant illumination of high color temperature. Different aspects about light transmission through various optical devices are discussed.

A color line scan camera is described in this paper which has three separate but simultaneous scans of Red, Green and Blue color with high resolution. Unlike previous cameras designed for color use in machine vision, this camera has high resolution per color (2098 pixels) without any non sensitive regions. This is achieved by employing a sensor which has three separate linear arrays integrated onto one chip with a different color filter over each linear array. The scan time between each color scan is equivalent to 8 scan lines. The camera produces an output per color which is digitized to 8 bits of grey scale at a total maximum speed of 30 MHz. This information can be obtained in a 24 bit wide format or the color outputs can be recombined in the camera so that each pixel is provided with three serial 8 bit values for each color. The camera is interfaced to a frame grabber which can correct for the line delay and provide display and analysis.

With increasing applications of image processing in manufacturing, it has become necessary to provide industry with powerful as well as flexible means of adapting the technology to the solution of different kinds of tasks at hand. Besides the standard components of an image processing system, the availability ofvarious software tools to fill the gap between design, digitization, measurement and the manufacturing phases should be emphasized. The users could e.g., desire to transfer grabbed images into certain computer graphics data format for further processing. They may want to apply image processing for the measurement of machine parts interactively or automatically, or they may want to convert the pictures processd, e.g. with edge detectors, into some CAD data format, or even to generate a numerical control file to machine the parts directly from the image processing data. With this motivation, two software packages FIPS—"Flexible Image Processing System", and IQIS— "Integrated QUality Inspection System" have been developed for manufacturing and quality inspection and have been used by industry for various purposes from simple image digitization and processing, luminance adjustment, data and file format conversion, generation of control files for numerical control in laser cuffing, to direct control of robot movement, and interactive as well as automatic machine part measurement.

The microelectronic industry is a protagonist in driving automated vision to new paradigms. Today semiconductor manufacturers use vision systems quite frequently in their fabs in the front-end process. In fact, the process depends on reliable image processing systems. In the back-end process, where ICs are assembled and packaged, today vision systems are only partly used. But in the next years automated vision will become compulsory for the back-end process as well. Vision will be fully integrated into every IC package production machine to increase yields and reduce costs. Modem high-speed material processing requires dedicated and efficient concepts in image processing. But the integration of various equipment in a production plant leads to unifying handling of data flow and interfaces. Only agile vision systems can act with these contradictions: fast, reliable, adaptable, scalable and comprehensive. A powerful hardware platform is a unneglectable requirement for the use of advanced and reliable, but unfortunately computing intensive image processing algorithms. The massively parallel SIMD hardware product LANTERN/VME supplies a powerful platform for existing and new functionality. LANTERN/VME is used with a new optical sensor for IC package lead inspection. This is done in 3D, including horizontal and coplanarity inspection. The appropriate software is designed for lead inspection, alignment and control tasks in IC package production and handling equipment, like Trim&Form, Tape&Reel and Pick&Place machines.

Inspection systems for wide web materials have been unable to effectively image fine defects as they are detected. The amount of data produced by highly parallel video inspection cameras can exceed 400 MBytes/sec. The system described in this paper is capable of analyzing and displaying a detected image within seconds of the event using a single frame grabber and a 386 computer. The system can operate at processing speeds of greater than 400 MBytes/sec since it makes use of a novel post processing algorithm within the camera itself. The video cameras are based on Time Delay and Integration technology to provide high grey scale resolution at high data rates and low light levels. The system has an adjustable resolution ranging from 256 to 24,000 pixels per line scanned. The scanning rate is adjustable to a maximum of 20,000 line scans per second.

This paper describes some texture-based techniques that can be applied to quality assessment of flat products continuously produced (metal strips, wooden surfaces, cork, textile products, ...). Since the most difficult task is that of inspecting for product appearance, human-like inspection ability is required. A common feature to all these products is the presence of non- deterministic texture on their surfaces. Two main subjects are discussed: statistical techniques for both surface finishing determination and surface defect analysis as well as real-time implementation for on-line inspection in high-speed applications. For surface finishing determination a Gray Level Difference technique is presented to perform over low resolution images, that is, no-zoomed images. Defect analysis is performed by means of statistical texture analysis over defective portions of the surface. On-line implementation is accomplished by means of neural networks. When a defect arises, textural analysis is applied which result in a data-vector, acting as input of a neural net, previously trained in a supervised way. This approach tries to reach on-line performance in automated visual inspection applications when texture is presented in flat product surfaces.

The automatic grinding of casted components handled by industrial robots gains increasing importance especially in connection with free formed surfaces. This paper describes an automatically working quality assurance cell that is able to test and evaluate the grinding results. Within this task image processing algorithms are used combined with simulated classifying neural networks to get at first a quality feature for the shape of free formed surfaces. The second step of the quality testing procedure is the search for surface flaws. They can be detected by adapted image filtering algorithms using polarized lighting sources. Another trained neural net supports this testing process, too. Herewith it is possible to detect and classify point and linear defects lying on the surface. By the examination of the scattering cone of infrared light sent to the surface microscopic defects such as roughness deviations off the target value are detected. These techniques have now been applied to armatures that are produced in the sanitary industry. The testing results are adapted to the needs of a control loop and transmitted to the grinding cell for a detailed touching up of the surface, if necessary.

This paper is based on resolving a problem found in the ceramics industry, namely Porcelanatto. Porcelanatto is a relatively new product, it is a new type of glazed tile and is used as a floor covering. Its manufacturing process consists basically of mixing two or three types of grains and after that the mixture is heated in an oven. Finally the mixture is polished and a perfectly flat surface is produced. Ideally, a piece of Porcelanatto should be a random mix of two or three types of grain where each grain has a fixed color and average size. Nevertheless, a selection of real tiles from the production line show slight differences in average grain size. These are responsible for the distinct appearance between tiles visible to the human eye once they have been placed on the floor. As it is impossible to obtain ideal tiles they must be classified into groups according to the criteria `appearance to the human eye' which is related to the average grain size. The description of an automatic system based on computer vision to classify them is the aim of this paper. The proposed system consists of: PC, image processing card with its library, black and white video camera CCD, illumination, data acquisition card and production line.

Links between the fractal Hausdorff-dimension, the Fourier transform of 2D scenes and image segmentation by texture are discussed. It is shown that the fractal Hausdorff-dimension can be derived by integration of the intensity of the spatial frequency domain (i.e. the Fourier plane) over a set of different band-limited spatial filters. The difference between a computational and optical approach to determine the Hausdorff-dimension are shown, with advantages of both methods discussed. Possible future directions of research/improvements are mentioned. Natural and simulated scenes are considered which apply to a wide range of situations in the agricultural and food industry.

This paper describes an application of image processing for the furniture industry. It uses an input data, images acquired directly from wood planks where defects were previously marked by an operator. A set of image processing algorithms separates and codes each defect and detects a polygonal approach of the line representing them. For such a purpose we developed a pattern classification algorithm and a new technique of segmenting defects by carving the convex hull of the binary shape representing each isolated defect.

This paper describes research into a non-contact system for the inspection of fine pitch electronic components on manufactured electronic assemblies. A far-field diffraction pattern from the leads of a tape automated bonding component is captured and stored on a frame store. The diffraction pattern is statistically represented and then classified using an artificial neural network. Results from simulation and experimentation show the feasibility of the technique.

The aim of the project is to develop a system for automatically inspecting the soldering of circuit boards. The system detects the quality of the soldering, determines if the board will work or not, and indicates the position of any incorrect soldering work. The system is based on the analysis of characteristics using a dedicated processor of 7 X 7 pixels. The method detects the quality of soldering work with a much quicker processing time than using conventional image treatment. The paper will describe the above mentioned system, the layout of the installation and the architecture of the board used as well as the algorithms implemented and the results obtained. The presentation will cover all areas of the project and will prove useful to those interested in automatic visual inspection techniques.

The visual inspection of many industrial products is based on contour analysis. The main task is the detection of defects like cracks and chips. The automation of this visual inspection procedure requires robust and flexible systems and algorithms. This paper describes new contour analysis algorithms for defect detection. Innovative techniques for contour filtering and analysis (e.g. scale-space analysis) will be presented. The performance of different approaches is shown by several industrial examples and applications.

The reliability of stand-alone machines and larger production units can be improved by automated condition monitoring. Analysis of wear particles in lubricating or hydraulic oils helps diagnosing the wear states of machine parts. This paper presents a computer vision system for automated classification of wear particles. Digitized images from experiments with a bearing test bench, a hydraulic system with an industrial company, and oil samples from different industrial sources were used for algorithm development and testing. The wear particles were divided into four classes indicating different wear mechanisms: cutting wear, fatigue wear, adhesive wear, and abrasive wear. The results showed that the fuzzy K-nearest neighbor classifier utilized gave the same distribution of wear particles as the classification by a human expert.

The overall performance of computer vision systems for automated visual inspection depends to a large extent on the quality of the lighting. This is especially the case when objects containing transparent and glossy parts have to be inspected, due to their special optical properties, which give rise to particular problems as well as possibilities. A typical aspect of this type of objects is, that the resulting image has a high dependency on the directional properties of the illumination, i.e. from where the light comes. This paper describes how a systematic search for the optimum lighting can be undertaken by making a classification and survey of the possible types of lighting that can be made with elementary shapes of light sources. For a series of different types of lighting created with these sources the resulting images for a typical object, chiefly transparent, are shown. In addition to the directional properties, some effects of polarization is treated. The images exhibit great differences both in general appearance and in ability to reflect the features to be inspected, whether these are shape, defects or contaminants. This shows, that there is a great variation in the possible ways to illuminate objects of this type, and it stresses the importance of optimizing the lighting to the particular task. Finally a case story with an object made from different materials, opaque and transparent, is treated. As is often experienced in practice, the best illumination results as a compromise between some more or less contradictory demands.

Modern machine vision systems won't require task specific hard- or software development in most cases. This means for the user short project realization times and minimized cost. Changes or new controls can easily be performed by a trained operator. Hard and software design of such a system is described in this paper, followed by two different applications, which were realized exclusively by menu-guided configuration of the built in set of control methods. The first is an in-line inspection of oven linings. To protect the following machine tool from being damaged, presence of holes and slits have to be controlled. Via serial communication, a host computer gets detailed information about the faulty item and the position of the defect on the lining. The second application describes an in-line inspection of brake pads. To ensure the lining is mounted to the correct side of the carrier, some characters on the carrier side have to be detected. According to the result the system triggers a turn over mechanism.

This paper describes the Artificial Intelligent techniques applied to the interpretation process of images from cast aluminum surface presenting different defects. The whole process includes on-line defect detection, feature extraction and defect classification. These topics are discussed in depth through the paper. Data preprocessing process, as well as segmentation and feature extraction are described. At this point, algorithms employed along with used descriptors are shown. Syntactic filter has been developed to modelate the information and to generate the input vector to the classification system. Classification of defects is achieved by means of rule-based systems, fuzzy models and neural nets. Different classification subsystems perform together for the resolution of a pattern recognition problem (hybrid systems). Firstly, syntactic methods are used to obtain the filter that reduces the dimension of the input vector to the classification process. Rule-based classification is achieved associating a grammar to each defect type; the knowledge-base will be formed by the information derived from the syntactic filter along with the inferred rules. The fuzzy classification sub-system uses production rules with fuzzy antecedent and their consequents are ownership rates to every defect type. Different architectures of neural nets have been implemented with different results, as shown along the paper. In the higher classification level, the information given by the heterogeneous systems as well as the history of the process is supplied to an Expert System in order to drive the casting process.

An optical sorter technology has been developed to improve profitability of a mine by using color line scan machine vision technology. The new technology adapted longers the expected life time of the limestone mine and improves its efficiency. Also the project has proved that color line scan technology of today can successfully be applied to industrial use in harsh environments.

The image of the froth surface in the mineral processing plant contains the information about the froth composition, which can be very useful for an automation of a technological process. The image processing and recognition method are the tools for analysis of this information. Different stages of technological process can be described by data extracted from optical images of processed materials. Images, transformed in the optical part of the hybrid system to the form suitable for feature extraction process, are read by the computer. The transformed data are analyzed in the digital part of the hybrid system. The set of image parameters, extracted from an image transforms, is evaluated by statistical analysis. The synthetic image descriptors (which are also the synthetic descriptors of different stages of technological process) are constructed on the base of training groups of images by discriminant analysis, during the learning step of presented methods. When image (and also technological process) descriptors are constructed, the method allows to recognize the technological process stage and next to control this process via the expert system. This paper presents the theoretical background of the method and the image analysis results.

A method for the experimental realization of optical correlation for pattern recognition by using phase only filters as a reference in a real-time joint transform correlator is presented. The intensity of the interferences produced by the optical Fourier transform of both scene and reference is registered by means of a CCD camera and displayed through a liquid crystal device. A second Fourier transform accomplishes a real-time correlation. In our implementation we have used a videoprojector as a spatial light modulator. Non linear effects in the joint power spectrum have been shown as a technique for improving discrimination in a joint transform correlator. The design of spatial correlation filters is another technique to enhance discrimination. Usually, these filters are computed in the frequency domain, although they can be carried out in object space, by computing their inverse Fourier transform, becoming the reference in a joint transform correlator. In this case a codification is necessary because of the negative values. Another problem involved in the implementation is the filter normalization because usually the filters have a lower transmission than the scene. In this paper, the normalization factor has been chosen in order to optimize the optical efficiency of the process.

In this paper we address the problem of the choice of visual primitives structure in multi- images analysis systems. We consider that this problem has a great consequences over the robustness of such systems, over their using in different vision tasks and over their computation time. Usually, methods used in image analysis systems (pattern recognition, 3D reconstruction, ...) tend to match visual primitives like pixels, regions or edges to recognize an object, to search candidate couples, and more generally to analyze multiple images. In this work we clearly demonstrate that the structure of such primitives is extremely important to obtain robust systems. We don't want to compare between different primitives, but between different structures that one of them can be represented. The primitive we have chosen, in this discussion, is the edge of objects. Finally, complete quantitative experimental results are shown with various indoor and outdoor real world senses. The system has been tested on both stereo pairs and images sequence, which are two different applications.

There is a growing demand for specific highperformance objectives. More and more often, certain specialized applications demand a quality close to the theoretical limits and also to manufacturing feasibility. To get such a sophisticated lens, the knowhow of all the fields concerned is called for. Optics development finds more and more complex solutions, particularly through the use of efficient computers and complex software for optimization. In many cases, the imaging quality is "near the diffraction limit". Therefore, the demands on design, development of technology and also, of course, on manufacturing of the optical components are continually rising. The limits of feasibility are very often touched or even overstepped

Micropropagation of plants is done by cutting juvenile plants and placing them into special container-boxes with nutrient-solution where the pieces can grow up and be cut again several times. To produce high amounts of biomass it is necessary to do plant micropropagation by a robotic system. In this paper we describe parts of the vision system that recognizes plants and their particular cutting points. Therefore, it is necessary to extract elements of the plants and relations between these elements (for example root, stem, leaf). Different species vary in their morphological appearance, variation is also immanent in plants of the same species. Therefore, we introduce several morphological classes of plants from that we expect same recognition methods.

The feasibility of a human face identification system using isodensity lines is demonstrated through experimental investigation. Isodensity lines are the boundaries of constant gray level areas obtained by quantizing a facial image. Utilization of these lines has the following advantages: (1) The technique is algorithmically and computationally simple to implement, both in hardware and in software. (2) Significant 3D structures can be reflected in the description of the face. (3) High discrimination accuracy is achieved, even for faces with spectacles or with thin bears (stubble). This system consists of a delineation of isodensity lines and a matching process. Extraction of the facial area is very easily put into practice by using a isodensity lines to represent the contour lines of the face. The matching process consists of template matching (global matching) and local compactness matching (fine matching) of registered and input isodensity lines. Experimental results show a 97.7% accuracy in matching 44 pairs of the same persons and a 100% accuracy in discriminating between 1892 pairs of different persons, including men with glasses or thin beards and women with or without make-up.

The Optical Radar System (ORAS) provides a complete set of data to derive 3D surface contours of bodies or of body-like scenes primarily in industrial environment. In addition, the intensity as back-scattered from the object is recorded in order to create combined (3D and conventional) images for special tasks in object characterization. ORAS as described in this paper is based on the time of flight principle. Its specifications, measurement results and applications are presented.

Scanning objects in an extended scene or large architectural structures, such as buildings, is an essential task used in numerous applications. In this paper we present the concept of a chirped laser radar with a range of 100 m and a design resolution of 3(DOT)10^-5 that exhibits an optoelectronic signal processing. The scanning signal is a 1 microsecond(s) long chirped pulse with 100 MHz bandwidth and is centered at 300 MHz. This signal is generated by a voltage controlled oscillator (VCO) that is driven by a direct digital synthesizer (DDS). The requirements imposed on the quality of the chirped signal are very high. These include temperature stabilization of the VCO and a programmed correct of its non-linear frequency response. The DDS facilitates the generation of a clean signal of the desired quality. The return signal is analyzed via an optoelectronic signal processor that measures the time delay between the target signal and the reference signal. The optoelectronic signal processor consists of two Bragg-cells and has two different output channels. Both channels evaluate the time delay of the two signals from the compressed pulses. In this paper we present the evaluation of the first channel, that measures this delay as a function of time. First experimental results indicate that the signal analysis of this channel gives a resolution of 5(DOT)10^-4. The evaluation of the second channel is not entirely completed yet. We present experimental results achieved with the radar using a single laser diode, which is intensity modulated by variation of the diode current, and scanning of co-operative targets.

The realization of a coherent frequency modulated continuous wave LIDAR aimed for the accurate measurement of short distances employing a distributed feedback tunable twin-guide laser diode is demonstrated. A single shot relative accuracy of 8(DOT)10^-5 (8 micrometers ) has been achieved at a distance of 10 cm. The results presented here prove the predictions calculated from theory.

The evaluations of limiting LFM-CW-Lidar's performances are obtained. The comparison of different ways of gas pollutions monitoring are represented. The probing lasers energetics and modulation unstability's and nonlinearity's affect are considered. The experimental setup's functional scheme is presented.

A simple device for small particles detection is presented, consisting of a single-mode diode or gas laser with external cavity. The basis of the measurement is the selfmixing that occurs in the laser active medium when the radiation, scattered back by the moving particle into the laser resonator, interferes with the field inside it and causes changes of a pump current. Information about the number, velocity and size of particles can be derived by processing the current pulses. This compact device is capable of detecting a single microndiameter particle movement.

A Lidar system designed by the authors to detect and observe the tracks of effluents emitted by ships at sea under propulsion by combustion of fossil fuels could have many useful applications. A prototype system uses a frequency-doubled Nd:YAG laser as the transmitter, at wavelengths of 0.53 and 1.06 micrometers , and examine the return for Rayleigh, Mie, fluorescence, and Raman scattering (hence called Advanced Rayleigh-Raman-Mie, ARRM), to determine thermodynamic and chemical conditions out to a given distance surrounding the Lidar operating station. We discuss the business potential of the system, its application to a number of technical and environmental problems, the potential for job creation by the use of such systems, and their value to both the users and the rest of society. One such application is discussed to a limited extent: Wide-area surveillance for fire detection, in both urban and rural applications. Conversations with the City of Huntsville's Fire Department have revealed that the system is a quantum leap in fire detection, reporting, and hence firefighting response. Constant motion through the sky of a beam of green light could be a public nuisance. Therefore in an urban setting use of a primary surveillance transmitter consisting of a frequency-doubled CO₂ laser is necessary for initial detection of the thermodynamic and fluid-dynamic indications of the rising plume, from the Mie and Rayleigh scattering. After each detection of a new plume, the system switches to the Nd:YAG transmitter for detailed characterization of the plume from the measured ARRM parameters.

An overview is given to the methods of laser spot displacement control for compensation of eye movements during the eye microsurgery operations. Two major techniques were experimentally investigated: (1) with a mirror controlled electro-mechanically, and (2) having acousto-optic Bragg cell with controlled spatial frequency of diffraction grating. In both cases, displacements of 2 mm from the point of sight fixation were realized with an accuracy 0.1 mm and 0.01 mm respectively. Attention is paid to the methods combining TV techniques with spatial coding of light irradiation for reflexes tracking.

Coherent laser radar receivers often generate electronic signals having very wide information bandwidths. 50 MHz bandwidths are routine and 2 GHz bandwidths are not uncommon. Processing wide bandwidth data to remove noise and to extract the required information can exceed the practical limits of electronic techniques. This paper presents the results of the development of an optoelectronic signal processing system for a coherent laser radar system. The system can accommodate signals up to 500 MHz having a duration as short as 17 microseconds or as long as 143 milliseconds.

A great number of articles, special reference books and monographs are dedicated to the problem of underwater objects visibility. However many important problems remain unsolved. It concerns the underwater objects visibility through rough surface mainly. The present article is dedicated to the main aspects of this problem, exactly to the underwater objects observation with the aid of certain devices using artificial (laser) illumination. The notion of the image signal/noise ratio is introduced which takes into account both shot noises and signal fluctuations caused by a random air/water interface. Various methods of image formation for optic observation systems are considered. It is shown that the best results are obtained with the aid of the system, which has the maximum space and time averaging of fluctuations during the signal formation. Most important is the conclusion that the equivalence of two most popular methods of image formation (known in the theory visibility in turbid media) is violated, when observation is performed through a rough sea surface. In the considered case the best parameters has a system, which uses a pulsed laser beam of illumination with a wide aperture and a multi-unit receiver with time gating. The examples of calculation of the main characteristics of such systems are given.

The spatial modulated wavefront conjugated laser signals structure destruction out of the isoplanatic zone theoretically and experimentally examined and the results of experimental investigations are cited. The method of laser signals structure disturbances diminishing is suggested. It based both on reemitted laser signal wavefront conjugation and amplitude spatial modulation of this signal in the plane of optical systems aperture. The spatial filters for arbitrary nontemporal structure of laser signals in object plane are offered. It was proved that some atmospheric lies exist along which there are no influence of atmospheric turbulence on spatial structure of laser signals. The efficiency of the suggested method was examined experimentally by Earth-lied atmospheric line of 500th meters length and IAG wavefront conjugated pulse laser. The possibility of amplitude and phase spatial structure of laser signals maintain and its parameters guidance out of the isoplanatic zone is confirmed.

An influence of phototarget illuminance on laser measuring systems resolving power is considered. The adaptive method of laser measuring systems resolving power correction in high illuminance conditions is suggested and investigated. The mathematical calculations and experimental investigations results are cited.

The PERCEPTRON LASER System is based on light detection and ranging (LIDAR) technology. Non-contact laser datacameras provide true 3D images, which eliminate depth ambiguities and make previously difficult or even impossible robot guidance applications feasible.

This paper discusses some areas of industrial applications of high accuracy laser rangefinders and laser rangefinding techniques that require noncontact survey and profile measurements to be made to points of diffusely reflecting objects that are difficult to reach. Structural layouts and characteristics are given for phase and pulsed laser rangefinders.

Methods of description and classification of facial expression based on isodensity maps are presented in this paper. Until now, almost all methods of recognition of facial expression have depended on the shapes and edges on facial elements (position of eyebrows, eyes, nose, mouth, etc.). However it is very difficult to extract such feature points and to describe the change of density variations such as wrinkles by using feature points above. In contrast to these methods, this paper adopts and defines isodensity maps that can represent density variations of facial images and level planes. In this method, isodensity map is described as a tree structure. As isodensity map changes appear as changes in the tree structure, facial expression changes can be detected by analyzing the changes in the tree structure for example, by Brother Splitting, Brother Merging, Father-Son Splitting, or Father-Son Merging.

Inspection is a very important task for quality control in a factory. The manual detection of surface defects is a time consuming and tedious task because defects are usually small and the area to be inspected is large. Image analysis is necessary to solve this problem. In this paper, morphological image processing for greylevel inspection is illustrated and applied on typical defects encountered on ceramics and textile material. A short introduction to mathematical morphology is also provided.

We present a complete procedure for automated inspection of engraved marks and ink marks placed on ophthalmic lenses. The engraved marks are small and weak, and are partially occluded by ink marks. The whole sense is noisy and has low contrast. A preprocessing for image enhancement is first performed. Then, a simplified numerical correlation algorithm is applied, resulting in a fast and accurate positioning of the marks. The procedure includes designing a model for the target, in a way that greatly reduces the volume of computed data. Three figures of merit are defined to ensure the robustness of the whole procedure: the contrast of the target on the background, the uniqueness of the correlation peak, and the similarity between target and model. By comparison of these figures against suitable threshold levels, decisions based on existence, multiplicity and completeness of the marks can be taken.

Tunnel constructions need regular safety checks. The measurement of the cross-sectional profile in relation to the actual track position is of particular importance in the surveys. It allows to monitor the observation of the required clearance gauge. Profile measurements inside Tunnels are also needed to determine suitable clearance for consignments exceeding the loading gauge. The deployment of vehicles with larger cross sections, e.g. double-decker carriages, means that clearances get tighter. This also means that demand made on technical surveying and measuring systems are becoming more stringent. Amberg Measuring Technique have developed an optical scanner for special use in tunnel surveys. The actual scanning device is a rotating mirror with scanning angle of full 360 degree(s) that allows the entire surrounding to be scanned without breaks in one test run. At each revolution of the scanning mirror, a cross-sectional tunnel profile, represented by 2500 distance values is recorded. Together with the profile data, the intensity of the reflected light is also recorded, representing a grey scale value for the surface brightness. High rotational speed of the scanning mirror allows rapid recording. The recording duration for 1 km of tunnel is about 10 minutes. The data are stored in digital form for subsequent processing on a computer workstation.