Computer Science Workbench. Editor: Tosiyasu L. Kunii

Transcription

1 Computer Science Workbench Editor: Tosiyasu L. Kunii

2 Computer Science Workbench N. Magnenat Thalmann, D. Thalmann: Image Synthesis. Theory and Practice. XV, 400 pp., 223 figs., including 80 in color B.A. Barsky: Computer Graphics and Geometric Modeling Using Beta-splines. IX, 156 pp., 85 figs., including 31 in color H. Kitagawa, T.L. Kunii: The Unnormalized Relational Data Model. For Office Form Processor Design. xm, 164 pp., 78 figs N. Magnenat Thalmann, D. Thalmann: Computer Animation. Theory and Practice. Second Revised Edition. xm, 245 pp., 156 figs., including 73 in color N. Magnenat Thalmann, D. Thalmann: Synthetic Actors in Computer Generated 3D Films. X, 129 pp., 133 figs., including 83 in color K. Fujimura: Motion Planning in Dynamic Environments. XIll, 178 pp., 85 figs M. Suk, S.M. Bhandarkar: Three-Dimensional Object Recognition from Range Images. xxn, 308 pp., 107 figs. 1992

3 Minsoo Suk. Suchendra M. Bhandarkar Three-Dimensional Object Recognition from Range Images With 107 Figures Springer-Verlag Tokyo Berlin Heidelberg New York London Paris Hong Kong Barcelona Budapest

4 PROF. MINSOO SUK Department of Electrical and Computer Engineering Syracuse University 121 Link Hall Syracuse, New York , USA PROF. SUCHENDRA M. BHANDARKAR Department of Computer Science University of Georgia 415 Boyd Graduate Studies Research Center Athens, Georgia , USA Series Editor: PROF. DR. TosIYASU L. KUNII Department of Information Science Faculty of Science The University of Tokyo Hongo, Bunkyo-ku Tokyo, 113 Japan ISBN-13: DOl: / e-isbn-13: Printed on acid-free paper Springer-Verlag Tokyo 1992 Softcover reprint of the hardcover 1st edition 1992 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in other ways, and storage in data banks. The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.

5 SERIES PREFACE Computer Science Workbench is a monograph series which will provide you with an in-depth working knowledge of current developments in computer technology. Every volume in this series will deal with a topic of importance in computer science and elaborate on how you yourself can build systems related to the main theme. You will be able to develop a variety of systems, including computer software tools, computer graphics, computer animation, database management systems, and computer-aided design and manufacturing systems. Computer Science Workbench represents an important new contribution in the field of practical computer technology. T08iyasu L. Kunii

6 PREFACE The primary aim of this book is to present a coherent and self-contained description of recent advances in three-dimensional object recognition from range images. Three-dimensional object recognition concerns recognition and localization of objects of interest in a scene from input images. This problem is one of both theoretical and practical importance. On the theoretical side, it is an ideal vehicle for the study of the general area of computer vision since it deals with several important issues encountered in computer vision-for example, issues such as feature extraction, acquisition, representation and proper use of knowledge, employment of efficient control strategies, coupling numerical and symbolic computations, and parallel implementation of algorithms. On the practical side, it has a wide range of applications in areas such as robot vision, autonomous navigation, automated inspection of industrial parts, and automated assembly. The major research emphasis in computer vision during the last few decades has been primarily on the extraction and analysis of information in intensity images. This was partly due to the lack of reliable range sensors. Today, range sensing technology has progressed to a point where fast, reliable, and economical range sensors are readily available. This has prompted a recent surge in researcb dealing with the processing and analysis of range data. The book intends to identify important issues in 3-D object recognition from range images and to describe recent advances on those issues. We have tried to present the material in the broader context of computer vision wherever possible. Computer vision draws from the rich and vast body of knowledge accumulated in the areas of image processing and artificial intelligence over the past several decades. Image processing algorithms such as edge detection, segmentation, and feature extraction and representation form the basis of low- and intermediate-level vision. Artificial Intelligence concepts and methodologies that deal with knowledge representation and problem solving via constraint propagation and constraint satisfaction provide the bases for high-level vision. Lowand intermediate-level vision algorithms deal largely with numerical, pixel-level data, whereas high-level techniques deal largely with symbolic entities. A typical vision system, such as one dealing with 3-D object recognition from range images, can therefore be considered to be a coupled system where numerical and

7 viii symbolic computation need to be closely and synergestically coupled in order to solve a given problem efficiently. Our discussion of the 3-D object recognition emphasizes this aspect of computer vision, namely the interaction between numerical and symbolic processing. The two most significant problems encountered during the process of 3-D object recognition, especially when dealing with multiple-object scenes with partial occlusion, are the combinatorial explosion of the search space of scene interpretations and the generation of sp.rious scene interpretations. We present a novel concept of using qtj(j/itative features to tackle these problems. The idea is to use qualitative features (symbolic) to control the combinatorial explosion (numerical) in a coupled system. We integrate all these components into an object recognition system that can recognize objects from multiple-object scenes with partial occlusion. Scenes containing polyhedral objects as well as complex curved objects are used for experiments. The magnitude of computation required in a typical computer vision system is huge. Fortunately, massively parallel computers are now widely available. Today, parallel computation is not only a topic for theoretical research, but also one with a wide range of practical applications. Keeping this in mind, the book describes in detail the parallel implementation of our 3-D object recognition algorithms on the Connection Machine and hypercube computers. The book consists of three parts that develop the subject matter in a natural and logical sequence. Part I describes the basic issues and concepts underlying computer vision, and 3-D object recognition in particular. The terminology is outlined, along with definitions of the basic ideas developed in the book. Range sensors and sensing technology, range image segmentation, feature extraction, and representation of features and models are reviewed. Popular recognition and localization techniques, including interpretation tree search, generalized Hough transform, matching of relational structures, and geometric hashing are then described. The shortcomings of existing vision systems that use these techniques are pointed out. Part II deals with the use of qualitative features in object recognition. We show how the use of qualitative features enables us not only to prune the search space, but also to improve the accuracy of recognition. This concept is demonstrated by experiments using multiple-object scenes of increasing complexity-polyhedral objects, objects made up of curved surfaces, and complex curved objects. Part III deals with performance and implementation issues. We present the sensitivity analysis of recognition and localization techniques and parallel implementation of our algorithms. This book is meant to address a fairly wide and diverse audience. It is suitable for a one-semester or a two-quarter graduate-level course in electrical engineering, computer engineering, or computer science. It could also be used as supplementary material for a graduate-level course in mechanical/aerospace engineering dealing with sensor-based or intelligent robotics. The book would provide excellent reference material for researchers in computer vision, pattern recognition, image processing, robotics, and artificial intelligence. Certain graduate students and researchers in biomedical imaging, remote-sensing and cartography may find some of the basic material herein very helpful.

8 IX We owe a debt of gratitude to many people for their help in preparing this book. We wish to extend our appreciation to Professor Tosiyasu Kunii, the Series Editor, for his encouragement, and Miss Yuka Hirahara of Springer-Verlag for her editorial help on numerous occasions and her patience. We are very grateful to R. Shankar, G. Ramamoorthy, M. Yang, Z. Xu, S. Choi, J. Koh, D. Kim, and A. Siebert. M. Suk gratefully acknowledges H. J. Lee for his help in using UTEXand Ms. E. Weinman for her careful reading of the manuscript. Finally, special appreciation to our family, Younghee, Eugene, Brian, and Swati. August 1992 Minsoo Suk Syracuse, New York S. Bhandarkar Athens, Georgia

9 Contents SERIES PREFACE.... PREFACE.... Credits for Copyrighted Material V VII xv 1 Introduction Computer Vision Three-Dimensional Object Recognition Representation Indexing Constraint Propagation and Constraint Satisfaction Common Goals of Three-Dimensional Object Recognition Systems Qualitative Features Study of Qualitative Properties in Low-level Vision Processes Qualitative Features in Object Recognition The Scope and Outline of the Book I Fundamentals of Range Image Processing and Three- Dimensional Object Recognition 15 2 Range Image Sensors and Sensing Techniques Range Image Forms Classification of Range Sensors Radar Sensors Triangulation Sensors Sensors based on Optical Interferometry Sensors Based on Focusing Techniques Sensors Based on Fresnel Diffraction Tactile Range Sensors Range Image Segmentation Mathematical Formulation of Range Image Segmentation Fundamentals of Surface Differential Geometry Surface Curvatures Range Image Segmentation Techniques

10 xii CONTENTS Edge-based Segmentation Techniques Region-based Segmentation Techniques Hybrid Segmentation Techniques 3.5 Summary Representation Formal Properties of Geometric Representations Wire-Frame Representation Constructive Solid Geometry (CSG) Representation Qualitative Representation using Geons. 4.5 Aspect Graph Representation EGI Representation Representation Using Generalized Cylinders. 4.8 Superquadric Representation 4.9 Octree Representation lO Summary Recognition and Localization Techniques Recognition and Localization Techniques-An Overview Interpretation Tree Search Hough Clustering Matching of Relational Structures Geometric Hashing Iterative Model Fitting Indexing and Qualitative Features Vision Systems as Coupled Systems Object-Oriented Representation for Coupled Systems Object-Oriented Representation for 3-D Object Recognition Embedding Parallelism in an Object-Oriented Coupled System Summary II Three-Dimensional Object Recognition Using Qualitative Features Polyhedral Object Recognition 6.1 Preprocessing and Segmentation Plane Fitting to Pixel Data Clustering in Parameter Space Post Processing of Clustering Results Contour Extraction and Classification Computation of Edge Parameters. 6.2 Feature Extraction Interpretation Tree Search Pose Determination

11 CONTENTS Scene Interpretation Hypothesis Verification Generalized Hough Transform Feature Matching Computation of the Transform Pose Clustering Verification of the Pose Hypothesis Experimental Results Summary XlII Recognition of Curved Objects Representation of Curved Surfaces Extraction of Surface Curvature Features from Range Images Recognition Using a Point-Wise Curvature Description Object Recognition Using Point-Wise Surface Matching Recognition Using Qualitative Features Cylindrical and Conical Surfaces The Recognition Process Using Qualitative Features Localization of a Cylindrical Surface Localization of a Conical Surface Localization of a Spherical Surface An Experimental Comparison Recognition of Complex Curved Objects Dihedral Feature Junctions Types of Dihedral Feature Junctions Matching of Dihedral Feature Junctions Pose Determination Pose Clustering Experimental Results Summary III Sensitivity Analysis and Parallel Implementation Sensitivity Analysis Junction Matching and Pose Determination Sensitivity Analysis Qualitative Features The Generalized Hough Transform The Generalized Hough Transform in the Absence of Occlusion and Sensor Error The Generalized Hough Transform in Presence of Occlusion and Sensor Error Probability of Spurious Peaks in the Generalized Hough Transform The Use of Qualitative Features in the Generalized Hough Transform

12 xiv CONTENTS Reduction in the Search Space of Scene Interpretations due to Qualitative Features Reducing the Effect of Smearing in Parameter Space using Qualitative Features The Probability of Random Peaks in the Weighted Generalized Hough Transform Determination of Pk(Z), Pk(Z) and P(k) : Weighted Generalized Hough Transform Parallel Implementations of Recognition Techniques Parallel Processing in Computer Vision Parallel Architectures Parallel Algorithms The Connection Machine System Organization Performance Specifications Object Recognition on the Connection Machine Feature Extraction Localization of Curved Surfaces Computation of Dihedral Feature Junctions Matching and Pose Computation Pose Clustering Object Recognition on the Hypercube Scene Description Model Data Scene Feature Data Pruning Constraints Localization Mapping the Interpretation Tree on the Hypercube Breadth-First Mapping of the Interpretation Tree Depth-First Mapping of the Interpretation Tree Depth-First Mapping ofthe Interpretation Tree with Load Sharing Experimental Results 273 BIBLIOGRAPHY 279 Index

13 Credits for Copyrighted Material Figure 2.1 Reprinted from: Besl, P.J. "Active Optical Range Imaging Sensors," in Advances in Machine Vision, Ed. Jorge L. C. Sanz, Springer Verlag, New York, 1989, pp. 1-63, (Figure 1.5, adapted). Copyright: 1989 Figure 2.2 Reprinted from: Besl, P.J. "Active Optical Range Imaging Sensors," in Advances in Machine Vision, Ed. Jorge L. C. Sanz, Springer Verlag, New York, 1989, pp. 1-63, (Figure 1.5, adapted). Copyright: 1989 Figure 2.3 Reprinted from: Jarvis, R.A. "A Perspective on Range Finding Techniques for Computer Vision, in IEEE Trans. Pattern Analysis and Machine Intelligence, Vol. PAMI-5, No.2, March 1988, pp , (Figure 15, adapted). Copyright: 1988 IEEE. Figure 2.4 Reprinted from: Besl, P.J. "Active Optical Range Imaging Sensors," in Advances in Machine Vision, Ed. Jorge L. C. Sanz, Springer Verlag, New York, 1989, pp. 1-63, (Figure 1.8). Copyright: 1989 Springer Figure 2.8 Reprinted from: Besl, P.J. "Active Optical Range Imaging Sensors," in Advances in Machine Vision, Ed. Jorge L. C. Sanz, Springer Verlag, New York, 1989, pp. 1-63, (Figure 1.9). Copyright: 1989 Springer Figure 2.9 Reprinted from: Inokuchi, S., Sato, K., and Matsuda, F., "Range Imaging System for 3D Object Recognition," in Proc. IEEE Inti. Con! on Pattern Recognition, Montreal, Canada, 1984, pp , (Figure 2). Copyright: 1984 IEEE. Figure 2.10 Reprinted from: Besl, P.J. "Active Optical Range Imaging Sensors," in Advances in Machine Vision, Ed. Jorge L. C. Sanz, Springer Verlag, New York, 1989, pp. 1-63, (Figure 1.14). Copyright: 1989 Springer Figure 2.11 Reprinted from: Leger, J.R. and Snyder, M.A., "Real Time Depth Measurement and Display Using Fresnel Diffraction and White Light Processing," in Applied Optics, Vol. 23, No. 10, May 15, 1984, pp , (Figure 1). Copyright: 1984 Optical Society of America. Figure 3.1 Reprinted from: Besl, P.J., and Jain, R.C., "Invariant Surface Characteristics for 3D Object Recognition in Range Images", in Computer Vision Graphics and Image Processing, Vol. 33, 1986, pp , (Figure 4). Copyright 1986 Academic Press.

14 xvi Figure 3.6 Reprinted from: Langridge, D.J., "Detection of Discontinuities in the first Derivatives of Surfaces", in Computer Vision Graphics and Image Processing, Vol. 27, 1984, pp , Figure 3(a). Copyright 1984 Academic Press. Figure 3.7 Reprinted from: Mitiche, A. and Aggarwal, J.K., "Detection of Edges Using Range Information" in IEEE Trans. Pattern Analysis and Machine Intelligence, Vol. PAMI-5, No.2, March 1983, pp , (Figure 1, adapted). Copyright 1983 IEEE. Table 3.1 Reprinted from: Besl, P.J., and Jain, R.C., "Invariant Surface Characteristics for 3D Object Recognition in Range Images", in Computer Vision Graphics and Image Processing, Vol. 33, 1986, pp , (Figure 5). Copyright 1986 Academic Press. Figure 3.10 Reprinted from: Jain, A.K. and Nadabar, S.G., "MRF Modelbased Segmentation of Range Images," in Proc. IEEE Inti. Conference on Computer Vision, Osaka Japan, Dec. 1991, pp , (Figure 2). Copyright: 1991 IEEE. Figure 3.11 Reprinted from: Jain, A.K. and Nadabar, S.G., "MRF Modelbased Segmentation of Range Images," in Proc. IEEE Inti. Conference on Computer Vision, Osaka Japan, Dec. 1991, pp , (Figure 4). Copyright: 1991 IEEE. Figure 3.12 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 3). Copyright: Pergamon Press. Figure 4.3 Reprinted from: Biederman, I., "Human Image Understanding: Recent Research and a Theory," in Computer Vision, Graphics, and Image Processing, Vol. 32, pp , (Figure 5, adapted), Copyright: 1985 Academic Press. Figure 4.13 Reprinted from: Chen, H.H. and Huang, T.S., "A Survey of Construction and Manipulation of Octrees" in Computer Vision Graphics and Image Processing, Vol. 43, 1988, pp , (Figure 1). Copyright 1988 Academic Press. Figure 4.14 Reprinted from: Chen, H.H. and Huang, T.S., "A Survey of Construction and Manipulation of Octrees" in Computer Vision Graphics and Image Processing, Vol. 43, 1988, pp , (Figure 2). Copyright 1988 Academic Press. Figure 5.2 Reprinted from: Ikeuchi, K. and Kanade, T., "Automatic Generation of Object Recognition Programs," in Proceedings of the IEEE, Vol. 76, No.8, August 1988, pp , (Figure 3). Copyright: 1988 IEEE.

15 xvii Figure 5.4 Reprinted from: Dhome. M. and Kasvand, T., "Polyhedra Recognition by Hypothesis Accumulation," in IEEE Trans. Pattern Analysis and Machine Intelligence, Vol. 9. No.3, May 1987, pp , (Figure 4). Copyright: 1987 IEEE. Figure 5.5 Reprinted from: Krishnapuram, R. and Casasent, D., "Determination of Three-Dimensional Object Location and Orientation from Range Images, in IEEE Trans. Pattern Analysis and Machine Intelligence, Vol. 11, No. 11, November 1989, pp , (Figure 1). Copyright: 1989 IEEE. Figure 5.8 Reprinted from: Wong, A.K.C., Lu, S.W. and Rioux, M., "Recognition and Shape Synthesis of 3-D Objects based on Attributed Hypergraphs," in IEEE Trans. Pattern Analysis and Machine Intelligence, Vol. 11, No.3, March 1989, pp , (Figure 3). Copyright: 1989 IEEE. Figure 5.9 Reprinted from: Stein, F. and Medioni, G., "Structural Indexing; Efficient 3-D Object Recognition" in IEEE Trans. PAMI, Vol. 14, No.2, Feb. 1992, pp , (Figure 5 (a) and (b), adapted). Copyright: 1992 IEEE. Figure 5.10 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 5). Copyright: 1991 Figure 6.1 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 1). Copyright: Pergamon Press. P\gure 6.2 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 2). Copyright: Pergamon Press. Figure 6.3 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 3). Copyright: Pergamon Press. Figure 6.4 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 4). Copyright: Pergamon Press. Figure 6.5 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 5). Copyright: Pergamon Press. Figure 6.6 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 6). Copyright: Pergamon Press.

16 XVlll Figure 6.7 Reprinted from: Bhandarkar, S.M. and Siebert, A., "Integrating Edge and Surface Information for Range Image Segmentation", to appear in Pattern Recognition, (Figure 7). Copyright: Pergamon Press. Figure 6.8 Reprinted from: Bhandarkar, S.M. and Suk, M., "Sensitivity Analysis and Pose Computation Using Dihedral Junctions," in Pattern Recognition, Vol. 24, No.6, pp , 1991, (Figure 1). Copyright: 1991 Pergamon Press. Figure 6.11 Reprinted from: Bhandarkar, S.M. and Suk, M., " Sensitivity Analysis and Pose Computation Using Dihedral Junctions," in Pattern Recognition, Vol. 24, No.6, pp , 1991, (Figure 5). Copyright: 1991 Pergamon Press. Figure 6.12 Reprinted from: Bhandarkar, S.M. and Suk, M., " Qualitative Features and the Generalized Hough Transform," to appear in Pattern Recognition, (Figure 6). Copyright: Pergamon Press. Figure 6.13 Reprinted from: Bhandarkar, S.M. and Suk, M., " Qualitative Features and the Generalized Hough Transform," to appear in Pattern Recognition, (Figure 7). Copyright: Pergamon Press. Figure 7.1 Reprinted from: Bhandarkar, S.M. and Suk, M., " Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 1). Copyright: 1991 Springer Figure 7.2 Reprinted from: Bhandarkar, S.M. and Suk, M., " Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 6). Copyright: 1991 Springer Figure 7.6 Reprinted from: Bhandarkar, S.M. and Suk, M., " Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 2). Copyright: 1991 Springer Figure 7.7 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 3). Copyright: 1991 Springer Figure 7.8 Reprinted from: Bhandarkar, S.M. and Suk, M., " Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 4). Copyright: 1991 Springer Figure 7.9 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 7). Copyright: 1991 Springer

17 XIX Figure 7.10 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 8). Copyright: 1991 Figure 7.11 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 9). Copyright: 1991 Figure 7.12 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 10). Copyright: 1991 Figure 7.13 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 11). Copyright: 1991 Figure 7.14 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 12). Copyright: 1991 Figure 7.15 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 13). Copyright: 1991 Figure 7.16 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 14). Copyright: 1991 Figure 7.17 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 15). Copyright: 1991 Figure 7.18 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 16). Copyright: 1991 Table 7.6 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Table 2). Copyright: 1991 Springer

18 xx Figure 7.19 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 17). Copyright: 1991 Figure 7.20 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 18). Copyright: 1991 Figure 7.21 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Applications, Vol. 4, 1991, pp , (Figure 19). Copyright: 1991 Figure 8.1 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 14). Copyright: 1991 Figure 8.2 Reprinted from: Bhandarkar, S.M. and Suk, M., "Recognition and Localization of Objects with Curved Surfaces," in Machine Vision and Applications, Vol. 4, 1991, pp , (Figure 15). Copyright: 1991 Figure 8.3 Reprinted from: Bhandarkar, S.M. and Suk, M., "Sensitivity Analysis and Pose Computation Using Dihedral Junctions," in Pattern Recognition, Vol. 24, No.6, pp , 1991, (Figure 6). Copyright: 1991 Pergamon Press. Figure 8.4 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Bough Transform," to appear in Pattern Recognition, (Figure 2). Copyright: Pergamon Press. Figure 8.5 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Bough Transform," to appear in Pattern Recognition, (Figure 3). Copyright: Pergamon Press. Table 8.1 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Bough Transform," to appear in Pattern Recognition, (Table 1). Copyright: Pergamon Press. Table 8.2 Reprinted fr:om: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Bough Transform," to appear in Pattern Recognition, (Table 2). Copyright: Pergamon Press. Table 8.3 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Bough Transform," to appear in Pattern Recognition, (Table 3). Copyright: Pergamon Press.

19 XXI Table 8.4 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Hough Transform," to appear in Pattern Recognition, (Table 4). Copyright: Pergamon Press. Figure 8.6 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Hough Transform," to appear in Pattern Recognition, (Figure 9). Copyright: Pergamon Press. Table 8.5 Reprinted from: Bhandarkar, S.M. and Suk, M., "Qualitative Features and the Generalized Hough Transform," to appear in Pattern Recognition, (Table 5). Copyright: Pergamon Press. Figure 9.1 Reprinted from: Tucker, L.W. and Robertson, G.G., "Architecture and Applications of the Connection Machine," in IEEE Computer, pp , August 1988, (Figure 1). Copyright: 1988 IEEE. Figure 9.2 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Figure 1). Copyright: 1992 Elsevier Science Publishers. Figure 9.3 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Figure 2). Copyright: 1992 Elsevier Science Publishers. Figure 9.4 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Figure 3). Copyright: 1992 Elsevier Science Publishers. Figure 9.5 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Figure 4). Copyright: 1992 Elsevier Science Publishers. Figure 9.6 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Figure 5). Copyright: 1992 Elsevier Science Publishers. Table 9.1 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Table 1). Copyright: 1992 Elsevier Science Publishers. Table 9.2 Reprinted from: Bhandarkar, S.M., "Parallelizing Object Recognition on the Hypercube", in Pattern Recognition Letters, Vol 13, No.6, June 1992, pp , (Table 2). Copyright: 1992 Elsevier Science Publishers.