Object Recognition Through Invariant Indexing

This book investigates how measures from a scene are incorporated into a recognition system. These measures are invariant, which means they do not change when an image is taken. This allows the system to recognize objects even if the image is changed or distorted.

Computer (robot) vision is a very challenging area of research. The problem of object recognition is central to computer vision in many applications such as the automatic sorting , selection, orientation, and inspection of manufactured items. It is also very important in navigation problems in mobile robots. In invarient indexing one computes measures from a scene that index into a base with a minimal search, so producing hypotheses of the identities of objects present in the scene. This book investigates how these measures are incorporated into a recognition system and also develops a range of projective indexes that can be used. The benefit of using projective measures is that they are unchanged by the imaging process. Review: This well-written monograph presents the author's research in object recognition through invariant indexing. This research work fills a void in theoretical and practical studies of computer vision. It can be used by researchers in computer object recognition and by implementors of object recognition systems. The length is just right. The presenation is clear. The index is adequate. All references are carefully selected and are cited in the text. This work is valuable from both theoretical and practical points of view, and I recommend it to readers interested in object recognition. Computing Reviews