No abstract

No abstract

This paper describes a method for using the high-level descriptions of objects (Le their models) to recognize them in an image. We view a complex object as a congregation of a set of component parts with simple shapes. Our model of an object, therefore, describes the shapes of its component parts, and states the geometrical relationships among those parts. It also includes a recognition strategy, which is a simple high-level description of how that object must be recognized. The shape descriptions of the parts are first used to extract a set of candidates for each part from the image. An object candidate is formed whenever a group of part candidates satisfy the model's geometrical relationships. A modelbased prediction and verification scheme is then used to verify (or refute) the existence of the object candidates with low certainty. This scheme not only substantially increases the accuracy of recognition, but also makes it possible to detect and recognize partially occluded and camouflaged objects. Another advantage of our approach is that to recognize a new object, we only need to define its model, and thus no programming is required. Moreover, the user's task is further simplified by the fact that each newly defined model is sufficient for recognizing a new category of objects.

No abstract