The primary aim of this book is to present a coherent and self-contained description of recent developments in three-dimensional object recognition from range data. The problem of three-dimensional object recognition, which deals with recognizing objects and estimating their poses from a range image, is one of both theoretical and practical interest. Due to the recent advances in range-sensing technology that have made range sensors fast, reliable and economical, the problem has acquired special importance in the computer vision research community. The book is unique in the sense that it provides an in-depth coverage of all important issues that pertain to the problem of 3-D object recognition in a single volume, from sensing to implementation. Following an up-to-date survey of range sensing and range image segmentation, the book discusses feature extraction and representation. Then the important issue of controlling/reducing the combinatorial complexity of the search space of scene interpretation is discussed in detail. In particular, it is shown that the use of qualitative features enables us to achieve the reduction of the search space as well as to improve the accuracy of recognition and localization. Finally, the sensivity of the recognition and localization is analyzed, and parallel implementation of them using Connection Machines and Hypercube computers is described.
