Introduction

Abstract

Three-dimensional object recognition, the subject of this book, is an important topic in computer vision. It concerns the recognition and localization of objects of interest in a scene from an image of the scene. The input image can be an intensity image (included in this category are color and multi-spectral images) or a range image. This problem is encountered in scenarios such as robot bin-picking, automated industrial inspection, autonomous navigation, to name just a few. In addition to the practical interest, three-dimensional object recognition is of considerable theoretical interest as well. It is an ideal vehicle for the study of the general area of computer vision. It involves typical 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. The intention of this book is to provide a detailed discussion of these and other important issues in three-dimensional object recognition, and to present some recent advances. The emphases are given to the dual issues of representation and control, and to the use of qualitative features to combat the combinatorial explosion of the search space of scene interpretations.