NATO ASI Series Advanced Science Institutes Series A series presenting the results of activities sponsored by the NA TO Science Committee, which aims at the dissemination of advanced scientific and technological knowledge, with a view to strengthening links between scientific communities. The Series is published by an international board of publishers in conjunction with the NATO Scientific Affairs Division A Life Sciences Plenum Publishing Corporation B Physics London and New York C Mathematical and Kluwer Academic Publishers Physical Sciences Dordrecht, Boston and London o Behavioural and Social Sciences E Applied Sciences F Computer and Springer-Verlag Systems Sciences Berlin Heidelberg New York G Ecological Sciences London Paris Tokyo Hong Kong H Cell Biology Barcelona Budapest I Global Environmental Change NATO-PCO DATABASE The electronic index to the NATO ASI Series provides full bibliographical references (with keywords and/or abstracts) to more than 30000 contributions from international scientists published in all sections of the NATO ASI Series. Access to the NATO-PCO DATABASE compiled by the NATO Publication Coordination Office is possible in two ways: -via online FILE 128 (NATO-PCO DATABASE) hosted by ESRIN, Via Galileo Galilei, 1-00044 Frascati, Italy. -via CD-ROM "NATO-PCO DATABASE" with user-friendly retrieval software in English, French and German (© WN GmbH and DATAWARE Technologies Inc. 1989). The CD-ROM can be ordered through any member of the Board of Publishers or through NATO-PCO, Overijse, Belgium. Series F: Computer and Systems Sciences Vol. 99 The ASI Series Books Published as a Result of Activities of the Special Programme on SENSORY SYSTEMS FOR ROBOTIC CONTROL This book contains the proceedings of a NATO Advanced Research Workshop held within the activities of the NATO Special Programme on Sensory Systems for Robotic Control, running from 1983 to 1988 under the auspices of the NATO Science Committee. The books published so far as a result of the activities of the Special Programme are: Vol. F25: Pyramidal Systems for Computer Vision. Edited by V. Cantoni and S. Levialdi. 1986. Vol. F29: Languages for Sensor-Based Control in Robotics. Edited by U. Rembold and K. Hormann. 1987. Vol. F33: Machine Intelligence and Knowledge Engineering for Robotic Applications. Edited by A K. C. Wong and A Pugh. 1987. Vol. F42: Real-Time Object Measurement and Classification. Edited by A K. Jain. 1988. Vol. F43: Sensors and Sensory Systems for Advanced Robots. Edited by P. Dario. 1988. Vol. F44: Signal Processing and Pattern Recognition in Nondestructive Evaluation of Materials. Edited by C. H. Chen.1988. Vol. F45: Syntactic and Structural Pattern Recognition. Edited by G. Ferrate, T. Pavlidis, A Sanfeliu and H. Bunke. 1988. Vol. F50: CAD Based Programming for Sensory Robots. Edited by B. Ravani. 1988. Vol. F52: Sensor Devices and Systems for Robotics. Edited by A Casals. 1989. Vol. F57: Kinematic and Dynamic Issues in Sensor Based Control. Edited by G. E. Taylor. 1990. Vol. F58 Highly Redundant Sensing in Robotic Systems. Edited by J. T. Tou and J. G. Balchen. 1990. Vol. F63: Traditional and Non-Traditional Robotic Sensors. Edited by T. C. Henderson. 1990. Vol. F64: Sensory Robotics for the Handling of Limp Materials. Edited by P M. Taylor. 1990. Vol. F65: Mapping and Spatial Modelling for Navigation. Edited by L. F. Pau. 1990. Vol. F66: Sensor-Based Robots: Algorithms and Architectures. Edited by C. S. G. Lee. 1991. Vol. F99: Multisensor Fusion for Computer Vision. Edited by J. K. Aggarwal. 1993. Multisensor Fusion for Computer Vision Edited by J. K. Aggarwal Department of Electrical and Computer Engineering The University of Texas at Austin Austin, TX 78712-1084, USA Springer-Verlag Berlin Heidelberg GmbH PPrroocceeeeddiinnggss ooff tthhee NNAATTOO AAddvvaanncceedd RReesseeaarrcchh WWoorrkksshhoopp oonn MMuullttiisseennssoorr FFuussiioonn ffoorr CCoommppuutteerr VViissiioonn,, hheelldd iinn GGrreennoobbllee,, FFrraannccee,, JJuunnee 2266--3300,, 11998899 CCRR SSuubbjjeecctt CCllaassssiiffiiccaattiioonn ((11999911)):: 11..22..99,, 11..22..1100,, 11..44..88,, 11..55..44 ISBN 978-3-642-08135-4 ISBN 978-3-662-02957-2 (eBook) DOI 10.1007/978-3-662-02957-2 TThhiiss wwoorrkk iiss ssuubbjjeecctt ttoo ccooppyyrriigghhtt.. AAllll rriigghhttss aarree rreesseerrvveedd,, wwhheetthheerr tthhee wwhhoollee oorr ppaarrtt o01f tthhee mmaatteerriiaall iiss ccoonncceerrnneedd,, ssppeecciifliiccaallllyy tthhee rriigghhttss o01f ttrraannssllaattiioonn,, rreepprriinnttiinngg,, rreeuussee o01f iilllluussttrraattiioonnss,, rreecciittaattiioonn,, bbrrooaaddccaassttiinngg,, rreepprroodduuccttiioonn oonn mmiiccrroofliillmmss oorr iinn aannyy ootthheerr wwaayy,, aanndd ssttoorraaggee iinn ddaattaa bbaannkkss.. DDuupplliiccaattiioonn o01f tthhiiss ppuubblliiccaattiioonn oorr ppaarrttss tthheerreeoofl iiss ppeerrmmiitttteedd oonnllyy uunnddeerr tthhee pprroovviissiioonnss o01f tthhee GGeerrmmaann CCooppyyrriigghhtt LLaaww o01f SSeepptteemmbbeerr 99,, 11996655,, iinn iittss ccuurrrreenntt vveerrssiioonn,, aanndd ppeerrmmiissssiioonn lfoorr uussee mmuusstt aallwwaayyss bbee oobbttaaiinneedd lfrroomm SSpprriinnggeerr--VVeerrllaagg.. VViioollaattiioonnss aarree lliiaabbllee lfoorr pprroosseeccuuttiioonn uunnddeerr tthhee GGeerrmmaann CCooppyyrriigghhtt LLaaww.. ©© SSpprriinnggeerr--VVeerrllaagg BBeerrlliinn HHeeiiddeellbbeerrgg 11999933 OOrriiggiinnaallllyy ppuubblliisshheedd bbyy SSpprriinnggeerr--VVeerrllaagg BBeerrlliinn HHeeiiddeellbbeerrgg NNeeww YYoorrkk iinn 11999933 SSooffttccoovveerr rreepprriinntt o01f tthhee hhaarrddccoovveerr 11ss tt eeddiittiioonn 11999933 TTyyppeesseettttiinngg:: CCaammeerraa rreeaaddyy bbyy aauutthhoorrss 4455//33114400 --55 443322 11 00 --PPrriinntteedd oonn aacciidd--fIrreeee ppaappeerr Preface This volume contains revised papers based on contributions to the NATO Advanced Research Workshop on Multisensor Fusion for Computer Vision, held in Grenoble, France, in June 1989. The 24 papers presented here cover a broad range of topics, including the principles and issues in multisensor fusion, information fusion for navigation, multisensor fusion for object recognition, network approaches to multisensor fusion, computer architectures for multi sensor fusion, and applications of multisensor fusion. The participants met in the beautiful surroundings of Mont Belledonne in Grenoble to discuss their current work in a setting conducive to interaction and the exchange of ideas. Each participant is a recognized leader in his or her area in the academic, governmental, or industrial research community. The workshop focused on techniques for the fusion or integration of sensor information to achieve the optimum interpretation of a scene. Several participants presented novel points of view on the integration of information. The 24 papers presented in this volume are based on those collected by the editor after the workshop, and reflect various aspects of our discussions. The papers are organized into five parts, as follows. Part I, Principles and Issues in Multisensor Fusion, covers the general issues in multisensor fusion. Since there is no broadly accepted definition of multi sensor fusion, these papers provide an interesting starting point for discussion. J an Olof Eklundh presents the importance of model selection for information inte gration with illustrative examples. James L. Crowley presents the general principles for integrating perceptual information into a coherent description of the world. J. K. Aggarwal and Chen-Chau Chu discuss the broad issues in multisensor fusion for computer vision and give an overview of recent results for scene segmentation and interpretation using images acquired via diverse sensing modali- VI ties. Demitri Terzopoulos surveys the physically based approach to data fusion using deformable models and simulated forces. Finally, Gerald T. McKee discusses a broad framework for fusion of information. Part II, Information Fusion for Navigation, presents four papers on fusing information for computer vision problems in navigation. Takeo Kanade et al. introduce a pixel-based algorithm that estimates depth and its uncertainty at each pixel. The algorithm combines Kalman filtering with iconic descriptions of depth that can serve as a general framework for low-level dynamic vision. Greg Hager considers a framework for describing sensor data fusion problems in terms of data representations, sensors, and sensing tasks, and its decision theoretical interpretation. He further presents an analysis of linear updating rules as well as a method of solving this class of problems. Hugh Durrant-Whyte describes techniques employed in geometric sensor integration from extracting initial de scriptions to integrating geometrically and geographically disparate sensory infor mation. G. Toscani and R. Deriche consider the problem of correspondence using token trackers for two similar tasks, calibration of a stereo set of cameras and fast motion detection. Part III, Multisensor Fusion for Object Recognition, presents experi mental results obtained from using multisensor fusion for object recognition. Uri Rembold et ai. discuss results on combining intensity and range image data for 3D object recognition. This timely research uses several range scanners to measure range and intensity simultaneously. Fusion is performed through the computation of edge and curvature information. In the second paper, Alberto Sanfeliu et ai. present a method for recognizing partially occluded three-dimensional objects from intensity images, using depth at a sparse set of points to reduce the number of candidate models in the initial hypothesis. X. Lebegue et al. discuss the use of color and geometric clues for recognition of man-made objects (concrete bridges) in outdoor scenes. The next paper, by R. M. Bolle et aI., describes a framework of VII fusion for visual recognition through the use of constraint satisfaction networks. In the final paper of this section, Amar Mitiche et al. compare methods for multi sensor object identification based upon statistical pattern classifiers, neural networks, and knowledge-based systems. This paper contains an extensive bibliography for the interested reader. Part IV, Computer Architectures for Multisensor Fusion, contains five papers on architectures for multi sensor fusion. First, V. Caglioti and M. Somalvico present a distributed architecture for fusing information that is capable of parallel execution of many activities. The second paper, by Andre Ayoun et aI., describes a general purpose architecture for multi sensor perception known as SKIDS - Signal and Knowledge Interaction with Decisional Control. Next, L. O. Hertzberger et al. describe algorithms for a SIMD processor array to achieve real time applications. Harry Stephanou et al. then present a classification algorithm using the theory of fractal sets in conjunction with the Dempster-Shafer theory. The final paper, by L. F. Pau et aI., presents a knowledge-based sensor fusion editor that uses a library of object-oriented methods and classes to achieve standard sensor fusion tasks. The papers in Part V, Applications of Multisensor Fusion, present five applications from diverse environments in which multisensor fusion for computer vision plays an important role. In the first, Rick Holben discusses the task of detecting man-made objects in a natural environment based on the analysis of image sequences. This computer vision system fuses thermal and video sensor data, and is oriented toward automatic surveillance applications. The second paper, by Paul Schenker et aI., reviews multi sensor techniques that are being used in space robotics. Application scenarios include telerobots for in-orbit spacecraft assembly and service, as well as autonomous roving vehicles for planetary exploration and sample collection. Next, Avi Kak et al. describe an approach to sensing strategies in a robot workcell with multi sensor capabilities. This system uses information from one sensor to develop a set of initial hypotheses and then selects the second sensing VIII modality that can best disambiguate the initial hypotheses. Sheldon Gruber presents a computer vision system that uses multiple sensors to measure the quality of surfaces. In this system, a hierarchical neural network that has been trained to recognize a selected set of machined surfaces is used to examine the sensor output. Finally, Clay Spence et al. detail results on the adaptive visuaVauditory fusion in the target location system of the barn owl. This paper highlights both the simplicity of such natural systems and the difficulty in understanding biological sensory systems. Altogether, these 24 papers cover a broad spectrum of topics and give a repre sentative picture of the current progress in multisensor fusion for computer vision among the leading research groups in Europe, Canada, and the United States. Finally, it is a pleasure to acknowledge the support of Norman Caplan of the National Science Foundation, whose active encouragement and support nourished the idea of a workshop on multi sensor fusion for computer vision from the initial proposal to NATO to its fruition at the Grenoble workshop. A number of individ uals contributed to the success of the workshop. Local arrangements were handled by Jim Crowley and Mirella Bello of LIFIA (IMAG), Grenoble, France. The workshop program committee consisted of Norman Caplan, Jim Crowley, Takeo Kanade (Carnegie Mellon University, U.S.A.), Alan Pugh (University of Hull, U.K.), and J. K. Aggarwal. Administrative details at The University of Texas at Austin were handled by J. K. Aggarwal and Debi Paxton. Most of all, it was the workshop participants who most contributed to its success, and special thanks are due to them for their enlightening and informative presentations at the workshop and in the following papers. Austin, Texas, October 1992 1. K. Aggarwal Table of Contents Preface .................................................. VII I. Principles and Issues in Multisensor Fusion Information Integration and Model Selection in Computer Vision . . . . . . . . . . . . . . 3 lan-Olof Eklundh Principles and Techniques for Sensor Data Fusion. . . . . . . . . . . . . . . . . . . . . . .. 15 lames L. Crowley The Issues, Analysis, and Interpretation of Multisensor Images. . . . . . . . . . . . . .. 37 I. K. Aggarwal and Chen-Chau Chu Physically-Based Fusion of Visual Data over Space, Time, and Scale . . . . . . . . . .. 63 Demetri Terzopoulos What Can Be Fused? 71 Gerard T. McKee II. Information Fusion for Navigation Kalman Filter-Based Algorithms for Estimating Depth from Image Sequences 87 Larry Matthies, Richard Szeliski, and Takeo Kanade Robust Linear Rules for Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . " 131 Greg Hager Geometric Sensor Fusion in Robotics (Abstract) . . . . . . . . . . . . . . . . . . . . . . . .. 151 Hugh F. Durrant -Whyte Cooperation between 3D Motion Estimation and Token Trackers (Abstract) . . . . . .. 153 G. Toscani andR. Deriche Three-Dimensional Fusion from a Monocular Sequence of Images . . . . . . . . . . . .. 155 I. L. lezouin and N. Ayache III. Multisensor Fusion for Object Recognition Fusion of Range and Intensity Image Data for Recognition of 3D Object Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 171 lianchi Wei, Paul Levi, and Ulrich Rembold Integrating Driving Model and Depth for Identification of Partially Occluded 3D Models ............................................... 195 A. Sanfeliu, M. Miaiios, and M. I. Dunjo x Fusion of Color and Geometric Information ........................... , 213 Xavier F. Lebegue, David C. Baker, and J. K. Aggarwal Evidence Fusion Using Constraint Satisfaction Networks. . . . . . . . . . . . . . . . . .. 239 Andrea Califano, Ruud M. Bolle, Rick Kjeldsen, and Russell W. Taylor Multisensor Information Integration for Object Identification . . . . . . . . . . . . . . . .. 255 A. Mitiche, R. Laganiere, and T. Henderson IV. Computer Architectures for Multisensor Fusion Distributing Inferential Activity for Synchronic and Diachronic Data Fusion. . . . . .. 279 V. Caglioti and M. Somalvico Real-Time Perception Architectures: The SKIDS Project . . . . . . . . . . . . . . . . . .. 293 A. Ayoun, C. Bur, R. Havas, N. Touitou, and J.-M. Valade Algorithms on a SIMD Processor Array. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 307 A. van Inge, L. O. Hertzberger, A. G. Starreveld, and F. C. A. Groen Shape and Curvature Data Fusion by Conductivity Analysis (Abstract). . . . . . . . .. 323 H. E. Stephanou and A. M. Erkmen A Knowledge-Based Sensor Fusion Editor. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 325 L. F. Pau, X. Xiao, and C. Westphal V. Applications of Multisensor Fusion Multisensor Change Detection for Surveillance Applications . . . . . . . . . . . . . . . .. 345 Rick Holben Multisensor Techniques for Space Robotics .................. ......... 367 P. Schenker, B. Wilcox, D. Gennery, and C. Anderson Coordinated Use of Multiple Sensors in a Robotic Workcell . . . . . . . . . . . . . . . .. 395 A. C. Kak, S. A. Hutchinson, C. H. Chen, S. N. Gottschlich, and K. D. Smith Neural Network Based Inspection of Machined Surfaces Using Multiple Sensors . .. 421 Sheldon Gruber Adaptive Visual! Auditory Fusion in the Target Localization System of the Barn Owl ............................................ 439 Clay D. Spence and John C. Pearson Index of Key Terms 451 VVorkshop Speakers 455
Description: