scientific article; zbMATH DE number 3972237

From MaRDI portal
Publication:3738604





The book presents a collection of papers on three-dimensional computer vision representing significant work by many of the major research groups in the field. Ten papers are based on talks given at the workshop on three-dimensional machine vision held in Washington D.C. in 1983 and an eleventh paper describes recent work at the University of Maryland. A first group of papers deal with range sensing and range data analysis. \textit{C. M. Bastuscheck} and \textit{J. T. Schwartz} from New York University describe in their paper Experimental implementation of a ratio image depth sensor (pp. 1-12) a variant of depth measurement by optical triangulation in which information is recorded simultaneously from an entire scene rather than point-by-point or plane-by-plane and present an experimental implementation of their method. \textit{O. D. Faugeras} and \textit{M. Hebert} from INRIA - France focus in their work The representation, recognition, and positioning of three-dimensional shapes from range data (pp. 13-51) on the task of constructing descriptions of static scenes containing various kinds of three-dimensional objects when range data and three-dimensional models of these objects are available and explore the problem of recognizing and positioning such objects in the work space. In order to represent models and scenes, they use linear primitives such as points, lines and planes. The paper of \textit{W. E. L. Grimson} and \textit{T. Lozano-Pérez} from M.I.T. (Model-based recognition and localization from sparse range data, pp. 113-148) discusses how local measurements of three-dimensional positions and surface normals may be used to identify and locate objects, from among a set of known objects. The objects are modeled as polyhedra having up to six degrees of freedom relative to the sensors. \textit{A. C. Kak}, \textit{K. L. Boyer}, \textit{R. J. Safranek}, and \textit{H. S. Yang} from Purdue University first show in their report Knowledge-based stereo and structured light for 3-D robot vision (pp. 185-218) that a knowledge based stereo using a single Marr-Poggio-Grimson channel is as effective for range mapping as a stereo that uses no a priori knowledge and all five channels; the second half of their report deals with structured light for three-dimensional vision, namely with a new scheme for extracting edges from range data and a high level symbolic processing program for robot vision. The paper of \textit{D. T. Kuan} and \textit{R. J. Drazovich} from A.I. \& D.S. - California (Model based interpretation of 3-D range data, pp. 219-230) describes a model-based approach for recognizing 3-D objects by using 3-D range data, where model objects are represented by a viewpoint independent volumetric model based on generalized cylinders. A second group of papers deal with stereopsis. \textit{D. B. Gennery} from California Institute of Technology describes in his paper Stereo vision for the acquisition and tracking of moving three-dimensional objects (pp. 53-74) a multiple-camera motion stereo solution for the acquisition and tracking of moving 3-D objects. \textit{W. E. L. Grimson} presents in his report Computing stereopsis using feature point contour matching (pp. 75-111) a version of the Marr-Poggio-Grimson computational model of the human stereo system, characterized as matching extended contours of feature points obtained from a range of different spatial frequency tuned channels. In his paper Representation and incremental construction of a 3-D scene model (pp. 149-183), \textit{M. Herman} from Carnegie Mellon University gives first the motivation for the approach of incrementally acquiring the scene model together with an overview of the system; then, the representation, construction and updating of the scene model are described, along with the task-specific knowledge used here; examples are shown of how urban scenes are reconstructed from complex aerial photographs. The third group of papers deal with model-based detection and recognition. \textit{V. J. Milenkovic} from Carnegie Mellon University presents in his paper Multiple resolution search techniques for the Hough transform in high dimensional parameter spaces (pp. 231-254), in order to find sets of parameters which fit the data best. \textit{L. G. Shapiro} from Virginia Polytechnic Institute discusses in her paper The use of numerical relational distance and symbolic differences for organizing models and for matching (pp. 255-270) both numerical and symbolic techniques for organizing a database of object models for fast access to the most appropriate model in the process of identification of an object extracted from an image. \textit{T. M. Silberberg}, \textit{D. Harwood}, and \textit{L. S. Davis} from University of Maryland present in their paper Three dimensional object recognition using oriented model points (pp. 271-320) a two stage algorithm that recognizes one or more 3-D objects in an image that contains the perspective projections of those objects. As it can be seen from this collection of works, computer vision techniques based on 3-D sensing and model matching are beginning to emerge from the laboratory and it is to be expected to make a major impact, especially in the area of robot vision. Although the aim of this book was not to cover all work done in the field - with one exception, all papers are from the USA - it could have been interesting to have some major works in the field done by authors from outside the USA, such as from Japan. Nevertheless, the book can serve as a useful reference for all those interested in this area.












This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3738604)