Theworld of the nuclear spins is a true paradise for theoretical and experimental physicists. It supplies, for example, most simple test systems for demonstrating the basic concepts of quantum mechanics and quantum statistics, and numerous textbook-like examples have emerged. On the other hand, the ease in handling nuclear spin systems predestines them for the testing of novel experimental concepts. Indeed, the universal procedures of coherent spectroscopy have been developed predominantly within nuclear magnetic resonance (NMR) and have found widespread application in a variety of other fields.

Optical methods are becoming interesting tools for high-resolution surface topography measurements. Micro- and macrostructure measurement techniques were developed to measure surface topography with resolutions of one micrometer or better. The progress made in the last few years is due to the development of the laser and solid state detector elements together with very powerful computer support for the processing of information. Methods for optical surface topography measurements based on the confocal principle and interferometry, together with future trends and some limitations, are discussed.

Today noncontacting optical 3D digitizers feature high sampling rates and are capable of obtaining highly resolved 3D images of object surfaces. In general these digital-range images contain a large amount of data and require powerful computers and large memories for further processing. It is shown that the number of data can be reduced by eliminating the redundant information contained in digital range images by applying pattern recognition techniques. Surface generation, contour lines generation, triangulation, and volume generation are explained for the different operation levels with respect to CAD/CAM.

In cooperation with the firm HILTI AG in Schaan the Swiss Federal Laboratories for Materials Testing and Research (EMPA) in Dubendorf, Switzerland is carrying out tests on masonry walls. The goal of these tests is to post-strengthen such walls in seismically endangered zones through the use of fiber-based composites. During the loading, the 3-D deformation behavior of the wall is determined. Each brick is provided with two measurement points. During the measurement phase the masonry wall deforms by up to 0.04 mm due to relaxation. Given a required deformation resolution of at least 0.04 mm, it is therefore necessary to employ a measurement technique that accurately and rapidly ascertains the wall condition in 3-D. In cooperation with the Institut fuer Photogrammetrie und Bildverarbeitung Technische Universitaet Braunschweig it was possible for the first time to employ five large- format Rollei LFC recording cameras synchronously for the deformation measurement. A spatial reconstruction of the wall shape was possible with an average standard deviation of 0.02 mm for the object coordinates in the plane of the wall, resp. 0.03 mm perpendicular to the wall.

Image acquisition is a prerequisite for photogrammetry. In close-range and especially in real- time applications, there is a trend to move from photographical recording to electronical image acquisition by means of CCD sensors. The main problem with these solid state semiconductor devices is still either a rather low resolution (2-D array sensors) or a long exposure time (line scanners). The performance figures currently achieved, and the pros and cons of the different scanning techniques and diversity of sensors used (full-frame CCDs vs. frame-transfer CCDs vs. interline-transfer CCDs) are discussed.

A digital photogrammetric system for automated 3D coordinate measurement in a production environment has been developed. For the image acquisition the Kodak DCS200 digital camera is used. This camera is based on a standard 35-mm camera. The results of the radiometric and geometric calibration of the DCS200 camera show the potential of this camera for photogrammetric applications. The software part of the system performs the detection, identification, and measurement of artificial targets present in digital images. These artificial targets are designed for automatic detection in images of a complex scene. For the identification of the targets a circular bar code is read by the image processing software. The least squares template matching method is implemented for the target image measurement. A precision better than 2% of a pixel was obtained for the target location. The 3D coordinate computation is performed by Geodelta's bundle adjustment package BINAER. It includes extensive statistical testing to assess the accuracy of the results. Tests with the DCS200 camera show a repeatability of 18 micrometer standard deviation on a test field 60 X 50 X 30 centimeter. The achieved precision is in the order of 2 (DOT) 10^-5.

The resolution of commercial digital sensors is a limiting factor for the achievable accuracy of photogrammetric on-line systems. The resolution can be increased, if the object is recorded sequentially with partial images orientated to each other. Partial images can be achieved by mounting a sensor on a bi-axial turning and tilting device. The geometric model for object reconstruction by means of a turning and tilting camera is introduced to a photogrammetric bundle adjustment. Systematic errors, such as the skew between the turning and tilting axis, can be taken into account.

The program PHAUST (Photogrammetrische Auswertestation fuer digitale Bilder) is a photogrammetric evaluation system for digital images. The applications of this program are the digitization of analog images, the processing of digital images, and the single or multi- image rectification in digital images. After a description of the system and an explanation of the single modules, the paper describes the possible applications of PHAUST, including some examples.

We discuss an off-line 3D measuring procedure under development in a project for the national program on machine vision in Finland 1992-1996. The procedure consists of four concurrent phases: (1) recording, (2) rectification, (3) digitizing, and (4) modeling. The recordings are based on sequential stereo videography, and the continuous object digitizing is based on the rigid normal case of stereography. The system modules are as follows: (1) two video cameras, (2) controllable feature projector, and (3) photogrammetric station. The procedure is interactively linked to a CAD/CAM-environment both for reverse engineering and quality control use.

In this paper the new active stereo vision head of the IPR is presented. It is designed to serve as a flexible sensing device mounted on a mobile robot platform. KASTOR has eight motorized optical and mechanical degrees of freedom. The article describes the design of KASTOR as well as the real-time vision system it is connected to. Then a new head calibration technique is presented. The perspective matrices are computed directly without the need to determine internal or external camera parameters. This is achieved by the observation of a reference object in a scene. As soon as the robot moves any of its degrees of freedom, the calibration has to be updated. However it is difficult and time expensive to define a reference object when the head is mounted on a mobile robot. A solution is to compute all possible perspective matrices in advance and to store them in the memory. During autonomous operation the correct matrices are selected according to the head-configuration. A standardizing technique is presented, which reduces the amount of matrices to be stored significantly. Finally the article discusses the quality of calibration when in use as active sensing device on a mobile robot.

In engineering geodesy a great variety of measurement tasks require contactless, high- accuracy, true-to-geometry measurements of nontargeted object features. For this reason a system using image processing is under development for automatic coordinate measurement. During the conceptual development phase of the instrument, it is very important to test the future hardware and software components of the system to achieve the best solution. This paper describes the investigation of the pointing accuracy of a computer-controlled motorized theodolite using an integrated CCD camera in its telescope. A program was developed to test the pointing accuracy within reach by means of a CCD camera and image processing software. Targets of different sizes and shapes were used at different distances for testing.

Abstract not available.

Configuring the setup for an industrial application of optical 3D sensing requires careful system design consideration. The camera and the frame grabber in particular determine the efficiency of the whole measurement system. The main aspect considered here is the synchronization of CCD cameras with a scanning (or flashlight) illumination system. Synchronization depends significantly on the transfer concept of the CCD sensor. The paper presents bases of CCD sensor concepts from the viewpoint of a user who has to configure an image processing system suitable for his special application. The performance of the digital image system in industrial application of in-process measurement of bending angles is investigated and the main sources of errors are sketched.

Camcorders are easy to use, transportable, provide a large integrated storage capacity, and are inexpensive. This makes them ideal image acquisition devices for many tasks in digital photogrammetry. Their geometric accuracy performance, on the other hand, is lower than that of comparable inspection type cameras. This paper gives a qualitative analysis of degrading effects such as the camera (color sensor and electronics) and video tape recording. Other practical issues, such as playback on another VCR, copying, and multiple playback, are also addressed. The geometric and radiometric characterization showed large response nonuniformities and degradations of the linearity. Geometric deformations depend on the recording/playback configuration and shear components exceeding 1 pixel were detected. The warm-up time was found to be similar to that of standard CCD cameras.

With the help of CCD images, realization of high-precision positioning and measurement has become the basic standard for machine vision and real-time photogrammetry systems. However, deformation and other degradation occurring during transmission are major limiting factors of the precision attainable with most current CCD cameras and framegrabbers. So precise radiometric and geometric transmission of images from CCD sensor to memory is a fundamental aspect of CCD camera calibration. A geometric calibration system that uses some image processing algorithms of the CCD camera based on the SOKKIA analytical plotter is discussed. Reliability is also discussed. The experimental results for the calibration of the CCD array are taken as an important quality index for CCD evaluation.

To effectively use automated zoom lenses for machine vision we need camera models that are valid over continuous ranges of lens settings. While camera calibration has been the subject of much research in machine vision and photogrammetry, for the most part the resulting models and calibration techniques have been for cameras with fixed parameter lenses where the lens' imaging process is static. For cameras with automated lenses the image formation process is a dynamic function of the lens control parameters. The complex nature of the relationships between the control parameters and the imaging process plus the need to calibrate them over a continuum of lens settings makes both the modeling and the calibration of cameras with automated zoom lenses fundamentally more difficult than that of cameras with fixed parameter lenses. In this paper we illustrate some of the problems involved with the modeling and calibration of cameras with variable parameter lenses. We then show how an iterative, empirical approach to modeling and calibration can produce a dynamic camera model of perspective projection that holds calibration across a continuous range of zoom.

A CCD camera and frame store provide image data that are affected by a number of imperfect processes. However, most of these can be quantified or estimated. For instance, the quantization process of allocating integer values from 0 to 255 for the intensity level of each pixel has well-known statistical properties. This paper discusses how these can lead to an estimate of the precision of location of target image centroids. Two centroid algorithms are analyzed. The theory is tested against experimental and simulated data.

Investigations have been carried out on the precision of target location in digital images to determine the influence of spatial and intensity quantization and threshold levels. Circular targets and centroiding methods are used in the study, and deviations from ideal images of the targets cause considerable loss of precision. Under favorable conditions, the precision of location was found to be better than 0.02 pixel size in both the x and y directions. The precision varies with several hundred percent according to the choice of the quantization parameters.

The knowledge of the precise step-response function of CCD sensors used as transducers is very important for displacement measurements. By the use of this function a subpixel accuracy can be achieved. However, the step-response function of a CCD sensor depends on the environmental parameters, the conditions of illumination, and therefore it is to be determined during rather than prior to the displacement measurements. This paper shows how precisely the step-response function has to be measured to achieve the highest accuracy. In addition, the authors investigate different interpolation methods to determine which step-response function fits best to the measured values.

Two separable approximation methods, Chebyshev approximation and SVD approximation, are described as ways of implementing an orientational filtering operation on a 2D image. They are both suitable for parallel implementation and enable the orientational filter to be performed on an N X N image in O(N²) time with a small amount of hardware. It is shown that these algorithms can be combined with Shensa's algorithm so that a multi- resolution pyramid can be efficiently constructed from orientational filter wavelets. Performance of the algorithms is evaluated by applying them to both Gabor filters and the first derivative of Gaussian filters. A VLSI architecture for orientational filters using the separable approximation is presented.

In digital photogrammetry, difficulties are often experienced when automatically matching differing viewpoints of similar targets, such as retro-reflective targets. In this paper, an alternative target matching technique to the epipolar line method is developed. The method, which is combined with a bundle adjustment process, is based on a 3D intersection and `epipolar plane,' as opposed to the 2D intersection of the epipolar line method. The theory of the approach is described and discussed. Comparisons are made between the epipolar line method and the 3D matching method. Results from both simulation and practical testing are given and critical evaluation is made of the technique.

For many applications in photogrammetry precise target localization is essential. Target localization by image processing is composed of several steps, from target detection to fine centering at the end of the process. The following paper describes a model for fine centering of a target based on Least Squares Image Matching (LSM). Since the image of a target has been convoluted by the image sensor LSM has to be extended by an appropriate mathematical model. Its derivation, together with practical results for fiducial marks (circles and crosses) and targets that are circular in object space, are presented.

The goal of the stereo algorithm is to determine the depth of objects from a stereo pair of images. The procedure of matching homogenous scenes from both images has been considered to be the most difficult problem. The approach described in this paper integrates the processes of feature matching and relational tree-matching techniques. Integration is necessary to ensure the high rate of correct matching results even in regions with many occlusions. The feature- matching technique used here is an adaptive one based on the boundary information from LoG operator, local invariant information detected by FFT operator, and the local curve signature. The relational tree technique is applied when the occlusion occurs. Experimental results are given for a typical photogrammetric stereo image.

The main task in photogrammetry is the reconstruction of 3D objects from 2D photographic or scanned images. Images must be orientated as a prerequisite for such a reconstruction. This paper presents a new method using spatial curves for object reconstruction and image orientation, complementing common methods based on distinct points. The first part of this paper deals with the representation of multi-dimensional free-formed curves by joined cubic polynomials. The interpolation and the parametrization of the representational curve are examined in detail. Additionally, observation equations of the adjusting curve are formulated. The second part of this paper is concerned with the application of the theory into the program system ORIENT. Finally two examples illustrate the process of image orientation and object reconstruction using this new method.

Presupposing relative orientation of projective bundles, the transformation to the normal case of `projective stereo pairs' is treated. The image coordinates do not refer to fiducial marks but to the content of the images whose image distances may be unknown. Hence the two images may originate from different nonmetric photographic cameras too. The method presented here uses projective image correlation and can be applied to both, digital and -- using photographic rectification -- analog images. The result is a normal case stereo pair without vertical parallaxes, which contains common parallel and horizontal epipolar lines.

Precise radiometric and geometric transmission of images from CCD sensor to memory is a fundamental aspect of CCD camera calibration. Line jitter and other degradations occurring during transmission are major limiting factors of the precision attainable with most current CCD cameras and framegrabbers. The influence on line jitter by framegrabbers with PLL (phase locked loop) and principal electronic components involved in synchronization and transmission are analyzed and discussed. The method and the algorithm for detection of the line jitter are discussed through the experiments detailed in this report. The experimental results of the detection of line jitter are taken as an important quality index for CCD evaluation.

This paper presents a new method using optical interferometry to measure the fractal dimension of the fault surface in metal material. It has been proved that the method can determine the dimension quickly and precisely. We explain the measurement principle and support a series of processing methods for fringe image.

Reverse engineering is the process of creating an engineering design model for existing parts or prototypes. We have developed a reverse engineering system where data is acquired with a scanning system that combines noncontact and contact digitizing methods. The noncontact sensor is a PC-based vision system that views the part from orthogonal orientations and captures the boundary points of the object. The images are then vectorized and a 2D CAD drawing of the part is created. The contact probe is mounted on a CNC machine, which is then guided by the NC code based on the 2D drawings of the part and captures the 3D coordinates of the points inside the boundaries of the object. The 3D coordinates are then used by the surface-modeling module of the system to create a 3D CAD drawing of the part, which is presented in a commercial CAD system. By combining vision sensing with contact probing we achieved speed and accuracy in the data extraction process. This paper describes the elements of the system and the CAD modeling procedure.

Abstract not available.

This paper describes a system for 3D data capturing in analytical close-range photogrammetry, which integrates all steps from orientation to on-line restitution in a CAD with superimposition. We explain the hardware and software concept of the Leica Photogrammetric Workstation, and show how the system meets the special requirements of close-range applications.

One of the basic conditions for reconstruction of structures is the determination of their real spatial form. This paper presents three applications of close-range photogrammetry, two of them using stereophotogrammetric digital image processing with the Digital Video Plotter Leica. In the third application of model measurement of the water surface deformation during the lock filling, an analytical solution was applied that uses the universal software ORIENT.

Three dimensional reconstruction of alpine terrain based on stereoscopic views from remote line scanners like SPOT requires robust and fast methods that find correspondences between homologue scene points in the stereo images. This matching of a large number of corresponding pairs of points is the most crucial step in the processing chain, both in computational effort and demands on robustness and accuracy. We propose a novel method coming from statistical pattern recognition that utilizes the advantages of hierarchical image representations. It is verified with a SPOT stereo pair. A comparison with the traditional gray- level correlation based matching philosophy shows better accuracy performance, while the decrease factor of necessary computational effort is of several orders of magnitude. A final discussion lists problems and tradeoffs that arise for the matching step, especially in alpine terrain.

The Digital Photogrammetric Station DPA based on a PC has been realized as a measurement system for close-range photogrammetry. Different digital image sources can be implemented and a large variety of image processing and measurement tools are available for 2D and 3D object reconstruction. In this paper performance and accuracy of the DPA measurement system are described and evaluated, giving some examples concerning industrial and architectural photogrammetry.

At the Institute of Navigation a new method of surveying historical sites and monuments has been developed and applied in field experiments. This method uses a 4D LaserScanner and sophisticated software to measure the 3D structure of the observed object and to partition and abstract the object's surface for automatic plan-drawing generation. By showing typical measurements and evaluation results, the application possibilities of the new method for conservation and restoration of historical buildings is demonstrated.

The use of digital photogrammetric close-range systems for on-line measuring tasks is of special importance in industry. A camera with motorized focus is advantageous. High- precision measurements are possible, if an optonumerical-controlled sensor displacement is applied. Special illumination and filter technique inside the camera is used for the sensor orientation by means of reference marks projected onto the sensor. Investigations have been carried out to optimize the image analysis for reference mark measurements.

The resolution of CCD sensors is strongly limited by pixel distance. For an application in high-accuracy position or displacement sensors it is very important to increase their resolution. The paper shows some methods to achieve a subpixel accuracy. The methods are tested by measurements.

This paper presents an automatic procedure of camera orientation developed for a digital close-range photogrammetric system. In this application, small bright balls mounted on a calibration frame serve as control points, since their shape in an image is invariant to the camera position and they are always imaged as circles. To recognize the circles in the image, an edge detection algorithm is exploited to extract the circular edges with subpixel accuracy. The circles are recognized by matching the shape of these edges with the shape of an ideal circular target. The central location of the circles and their diameters can be determined from these edge points. The determination of the identification of the circles is a problem of artificial intelligence. The list of the circles in the image must be arranged in the order of the balls in 3D world. A fast search is described that is based on exploiting the available information in order to limit the number of possible alternative orders of the targets. In this way, the search can be achieved efficiently. The identification of circles results in the correct numbers being attached to the corresponding circles. Finally, the precise camera parameters are calculated by bundle adjustment.

An approach is described for automatic relative orientation of a stereo pair of digital aerial images. Feature-based and area-based image-matching techniques are combined to automate the procedure of determination of conjugate points in the images. A coarse-to-fine strategy (image pyramids) and robust estimation techniques are incorporated into the approach. It was developed as one of the automation-oriented software components of a digital photogrammetric workstation. Results obtained from practical examples are presented and evaluated.

The projective transformation of 3D object points to 2D image points can be expressed in homogeneous coordinates by the 3 by 4 direct linear transformation (DLT) matrix. If the object points are coplanar, the transformation can be expressed by the 3 by 3 DLT matrix. This article presents non-iterative, fast and exact methods for extracting the physical camera parameters from the 3 by 3 matrix. Three parameters must be known or estimated in order to extract the remaining eight. The decomposition problem is solved under three different assumptions about known parameters. All three methods are based on solving two quadratic equations in two unknowns. The equations originate from the orthogonality constraints on the first two columns of the rotation matrix. The methods are well suited for finding initial values in point based camera calibration. They solve the coplanar perspective-n-point problem for any n >= 4.

For the examination of mixing processes in turbulent flows a system based on a high-speed solid state camera has been implemented, which allows for the quasi-simultaneous acquisition of sequences of tomography data in flows. In these data, velocity fields are being determined by 3D least squares matching. The paper shows the hardware configuration for the flow tomography data sequence acquisition and the algorithm for 3D least squares matching with additional constraints for tracking in tomography voxel data.

We present a double-heterodyne interferometer for absolute distance measurements at rough surfaces up to 100 meters. The two wavelengths are generated by frequency shifting the light of a monomode laser diode using a 500 MHz acousto-optic modulator. The synthetic wavelength obtained is (Lambda) equals 60 cm. In order to yield unambiguity up to 100 meter distance two measurements have to be taken using slightly different frequency shifts of 500 resp. 501.5 MHz. Measurements showing a resolution of 0.1 mm are presented.

Projection of coded patterns is a popular method for the acquisition of dense range images. Usually, the binary code or gray code is used for projection patterns. In this paper we investigate the light plane identification error in the two codes. Our main result is that the binary code and gray code have the same error characteristic while the partial gray code is superior to the partial binary code.

This paper describes a new method for fast 3D measurement based on active triangulation. In this approach the scene is inspected by a typical CCD camera and illuminated with a single structured light consisting of a multiple circular formed pattern. Data acquisition with video frame rate is guaranteed by the use of special picture preprocessing hardware and the reading of off-line computable table entries that represent 3D information. Based on a high-accuracy camera model and subpixel resolution this new method performs 3D measurement with an average accuracy of 2^*10^-4 relative to depth of field. Therefore this method is suitable for real-time industrial applications.

This paper describes the concept of differential inspection using a digital 3D imaging laser camera. Shapes can be compared and analyzed to a resolution smaller than 25 micrometers . Differential information is displayed as a high-definition image, color-coded for human interpretation. The same concept of differential inspection can be applied to 3D color images in order to monitor shape and color variations. Potential applications are discussed.

The instrumental errors of an analytical plotter derive mainly from the stage of the analytical plotter, so the stage calibration is a very important quality index for evaluating the measurement accuracy of the analytical plotter. With the help of CCD images, realization of high-precision positioning and measurement has become the basic standard for machine vision and real time photogrammetry systems. The software system, which employs some image processing algorithms for the automatic stage calibration with the CCD camera, based on the analytical plotter which SOKKIA researched and developed, is discussed. The reliability and validity are also discussed.

This paper presents the frequency-fixing pulse width modulation and adjustable microstep- driven system of a step motor controlled by the MCS-51 single-chip computer. The system uses single-voltage driving. The proprietary software is configured in the system. On the same supporting hardware, the numbers of the microstep can be changed from zero to 255 by operating the keyboard. The rotating speed, direction, and displacement can be controlled directly. The system applies the technology of the digital frequency-fixing PWM and the feedback control, of the microcurrent for electromagnetic noise elimination and low power losses. The curve of the experiment proves that the system has good quality.

We analyze in this paper the relationship between geodetic measurement instrumentation, measurement processes, and man-hours required. To initialize this research, we carried out a detailed study of the current practice using standard tools and techniques from `industrial engineering.' We ran an experiment with geodesy students to carry out a time and motion study of the one-man-system of geodimeter and compared it with a traditional `two-man- setup.' The time analysis was done by videotaping the user during the operation and simultaneously recording the time of keystrokes in the recording unit of the instrument. We found that the students lost 15% to 50% for setting up the instrument and storing various initial information. These times are the same for the conventional two-man setup. We also determined average times of different parts of setup and measurement operations.

In many automated reel stocks, the storage capacity has been increased using a special kind of vacuum load suspensor, which is able to pick up a stored reel without touching the exterior face. To facilitate the suspension process, this load suspensor includes a conical centering pin that enters the reel core at the end of the downward movement. With this paper we present a 3D videometric measuring system, that contains a camera mounted axially inside of the centering pin. Thus, the system is able to measure the lateral displacement between the centering pin and the core of the reel to be picked up. It supports a correct centric load suspension in order to avoid, damages to the reel by the centering pin at the end of the downward movement. It is also able to correct the crane position before a hanging load is released, to minimize the allowed lateral distance between adjacent stacks.

The paper presents a method of determining 3D displacements of points, called controlled points, using theodolite and reflections from flat mirrors. Displacement components (dx), (dy), (dz) are determined by means of observed increases of horizontal directions ((Delta) (alpha) ) and vertical directions. ((Delta) (beta) ).

In this paper a method is proposed to generate hypotheses about the existence of man-made objects from an aerial photograph. These hypotheses are used as the initial estimates for an estimation procedure that results in a volumetric scene description. The hypothesizing method is based on relaxation and bipartite matching. Its input consists of a description of the regions in the image, as found by a segmentation procedure. The output consists of 3D primitive objects.

The reconstruction of a surface having already matched corresponding points from stereo images (disparities) is a nontrivial task. This paper presents a new technique, the so-called Locus method, that exploits sensor geometry to efficiently build a terrain representation from stereo disparities. The power of this approach is the efficient and direct computation of a dense elevation map in arbitrary resolution. Additionally it proposes to solve problems like occlusions, ambiguities, and uncertainties caused by stereo matching errors. We extended the Locus method for active range finder data to the stereo disparity mapping case. For this reason, a newly developed fast matching method is utilized that provides dense disparity maps, hence a disparity for each input pixel. Once this data set is given, the Locus method can be applied in a straightforward and efficient way to gain a robust 3D reconstruction of the observed surface. It operates directly in image space, using dense and uniform measurements instead of first converting to object space. Experiments on synthetic and natural environment data show that the Locus method is less sensitive to disparity noise than traditional reconstruction.

This paper describes techniques for two (or more) camera geometry structured illumination range finders that eliminates most spurious range values that arise from specular reflections from shiny (e.g. metallic) parts. The key observation is that specular reflections produce range values that depend on camera position. Hence, by using two cameras, several consistency tests can be applied that eliminate most spurious range data. This paper describes the constraints that underlie the tests and shows sample performance.

This paper investigates the accuracy potential of object-based least squares matching for the automatic derivation of digital terrain models (DTMs). Some signal-theoretic-based improvements were made for sampling in the object space. Several sampling rates were tested in combination with different amounts of noise. The investigations were carried out for image strips from 3-line cameras. Most results hold for frame images as well. A few procedures were investigated for reducing the computation time and the resulting effect on the accuracy. It is shown that the computation can be accelerated by nearly a factor of 10 without any significant loss of accuracy.

This paper introduces the setup of the Remote Coordinate Measuring System (RCMS), which uses two T3000 informatic theodolites. The system data recording module is on-line storage but off-line operation. Angle measurements form two T3000s are transmitted automatically to a Wild GRE4, which store the observational data for later analysis. Using the RCMS, authors check the dead-weight deformation of a revised Casse-Grainian antenna, with a 25-inch diameter. The results indicated that the RCMS can monitor the dead-weight deformation when the antenna is in certain positions.

A prerequisite to the flexibility of machine vision systems is the ability to design a configuration of sensor stations that will provide for accurate triangulation for each new inspection application. Because of the number of design considerations and constraints on sensor station placement, and the complexity of interrelations between them, this task is currently carried out by experienced photogrammetrists, i.e. experts. In this paper we review progress in examining the feasibility of an expert system solution to the automation of this design task.

In this paper we compare the results of DEM estimation by a photogrammetric operator with the analytic plotter LEICA AC3 and an automatic object space based reconstruction method (Facets Stereo Vision). We are working with close-range images of natural stone surfaces, taken in cooperation with curators of monuments to monitor the weathering of stones. Two windows representative of the surface were selected and DEMs evaluated by an operator and Facets Stereo Vision. For the computation with Facets Stereo Vision a horizontal plane was introduced as initial value. Image pyramid techniques were applied for the final DEM estimation. Differences, problems of the comparison, and the results are discussed.

This paper describes a 3D vision system that allows the derivation of gripping information for a robot. Its capabilities are demonstrated in a benchmark that consisted in the task of clearing cafeteria trays. The system can handle a fair number of objects with substantial mutual occlusion. The objects can be transparent and specularly reflecting. In order to guarantee true 3D object recognition the system used three cameras. Two of these take `skylines' of the tray with dishes and cutlery pieces on it from the side, and a third camera takes a gray-level image from above. A ROI in each of the skyline contours pertaining to the highest object is analyzed, and features are extracted. Information from all three images is combined in a rule-based evidence accumulation scheme to determine the identity of the object in focus. The same features that voted for this object also allow the determination of its position and orientation. Finally the gripping points are calculated, depending on object, pose, and approach direction.

Measuring image motion and inferring scene geometry and camera motion are main aspects of image sequence analysis. The determination of image motion and the structure-from-motion problem are tasks that can be addressed independently or in cooperative processes. In this paper we focus on tracking prominent points. High stability, reliability, and accuracy are criteria for the extraction of prominent points. This implies that tracking should work quite well with those features; unfortunately, the reality looks quite different. In the experimental investigations we processed a long sequence of 128 images. This mono sequence is taken in an outdoor environment at the experimental field of Mercedes Benz in Rastatt. Different tracking schemes are explored and the results with respect to stability and quality are reported.

The paper describes a CCD TV-camera-based system for real-time multicamera 2D detection of retro-reflective targets and software for accurate and fast 3D reconstruction. Applications of this system can be found in the fields of sports, biomechanics, rehabilitation research, and various other areas of science and industry. The new feature of real-time 3D opens an even broader perspective of application areas; animations in virtual reality are an interesting example. After presenting an overview of the hardware and the camera calibration method, the paper focuses on the real-time algorithms used for matching of the images and subsequent 3D reconstruction of marker positions. When using a calibrated setup of two cameras, it is now possible to track at least ten markers at 100 Hz. Limitations in the performance are determined by the visibility of the markers, which could be improved by adding a third camera.

To get the position and orientation of a moving object, single images of the object supplied with fixed markings are taken by a CCD camera. The camera is included in a video-theodolite to obtain a nearly unlimited field of view and to obtain the exterior orientation each time. The trajectory of the object can be determined with an image sequence. This paper describes the mathematical formulation for the determination of orientation and position of the object. The problem is based on an inverse formulation of the resection in space, as the orientation and position of the camera is known. Furthermore an image processing algorithm is described to extract and match the control points within the image. The results of a test measurement of a moving object on rails are shown together with the accuracies and measuring frequencies achieved.

A new approach for photogrammetric measurement of deformation based on the known proposals of Facet Stereo Vision (Wrobel, Ebner, et al.) is presented. The approach of FAST- vision defines independent values of the object coordinate Z and the optical density D at discrete positions X, Y in the object space. These unknowns can be determined by a least square matching procedure together with the parameters of the interior and exterior orientation of the camera. In our approach the unknown displacement can be calculated together with the other unknowns of the FAST-Vision approach. Only gray values of the digital images are considered as observations. Therefore a number of different loading states can be determined in one least square adjustment process. This approach has been adapted to very small object surfaces (approximately equals 15 * 15 mm). The arrangement consists of two CCD-cameras (Kodak- Videk-Megaplus), which were fitted at a mechanical testing machine. The procedure is demonstrated by an example from fracture mechanics.

The basic steps of automatic surface reconstruction from digital images are briefly outlined, with attention to the factor determining the final quality of the process. The reconstruction of the deformation surface of a metal plate stressed to failure is then presented. The automatic search of the interesting points, the identification of conjugate pairs, and the least squares matching are described, outlining limits and characteristics peculiar to the chosen method. The results are also discussed in comparison with a previous photogrammetric survey by analog imagery.

Attached-grating moire was used to measure the displacement field of a wall that was sheared with both constant and cyclic load and, in a subsequent experiment, was deformed by a perpendicular load. A moire grid of large size (2 X 3 m²) was realized on the object surface. With its aid, the shear component in plane and one component of the out-of-plane displacement field, respectively, were determined by a phase shift algorithm. The results are compared to photogrammetric 3D data that were taken simultaneously. The calibration is discussed. We sketch the influence of phase errors and propose an algorithm to correct for them.

Monitoring the translations of points on structures and machinery is an important procedure in industrial engineering. Two low-cost translation monitoring systems using electronic tachometers have been developed and tested at the University of Calgary. The systems use both network and nonnetwork (direct) methods to measure relative point translations with submillimeter accuracy. The systems are simple and efficient, requiring only one observer using normal surveying instruments. Two case studies undertaken by the authors are described, demonstrating applications of the new monitoring systems. The first study is the measurement of thermal growth of a rotating gas turbine/compressor system during startup. The thermal movements were measured and found to be consistent with their expected behavior to a precision of 0.2 mm. The second study is the measurement of point displacements on a concrete structure and test frame during a load test at the University's Civil Engineering Structures Lab. The translations measured agree with the theoretical values from a finite element analysis.

The introduction of vision metrology (VM) systems into various facets of industrial measurement is slowly gaining momentum. Research and development attention is centered on multisensor stereo configurations with near real-time measurement feedback. The potential of single-sensor VM has been largely overlooked, in spite of its promise as an automated on-line high-accuracy 3D measurement tool in large-scale manufacturing. In this paper general attributes of single-sensor VM systems are discussed and two operational scenarios are outlined, one of which involves integration with traditional off-line optical triangulation systems. Example applications of single-sensor VM in the aerospace, nuclear, and automobile industries are also briefly summarized.

The possibility for testing and calibration of industrial robots becomes more and more important for manufacturers and users of such systems. Exacting applications in connection with the off-line programming techniques or the use of robots as measuring machines are impossible without a preceding robot calibration. At the LPA an efficient calibration technique has been developed. Instead of modeling the kinematic behavior of a robot, the new method describes the pose deviations within a user-defined section of the robot's working space. High- precision determination of 3D coordinates of defined path positions is necessary for calibration and can be done by digital photogrammetric systems. For the calibration of a robot at the LPA a digital photogrammetric system with three Rollei Reseau Scanning Cameras was used. This system allows an automatic measurement of a large number of robot poses with high accuracy.

The paper outlines the application and development of LEICA's triangulation and tracking systems to generate the reference 3D data needed for the static calibration of industrial robots, such as those used in automobile manufacture. Improvements in static positional accuracy are, for example, critical to the success of off-line robot programming, and some successful test measurements have been achieved with the motorized theodolite system, SPACE. Knowledge of a robot's dynamic performance is also required before it can be improved. Since the theodolite system is not suited to dynamic measurement, use of the interferometric laser tracking system SMART has also been investigated. This can potentially solve both static and dynamic tasks and could offer a more universal solution. Developments are less advanced here, but initial results are encouraging.

Optical tooling methods used in industry are rapidly being replaced by new electronic sensor techniques. The impact of new measuring technologies on the production process has caused major changes on the industrial shop floor as well as within industrial measurement systems. The paper deals with one particular industrial measuring system, the manual theodolite measuring system (TMS), within the aircraft and automobile industry. With TMS, setup, data capture, and data analysis are flexible enough to suit industry's demands regarding speed, accuracy, and mobility. Examples show the efficiency and the wide range of TMS applications. In cooperation with industry, the Video Theodolite System was developed. Its origin, functions, capabilities, and future plans are briefly described. With the VTS a major step has been realized in direction to vision systems for industrial applications.