RECOGNITION USING REGION CORRESPONDENCES

Recognition systems attempt to recover information about the identity of observed objects and their location in the environment. A fundament al problem in recognition is pose estimation. This is the problem of u sing a correspondence between some portions of an object model and som e portions of an image to determine whether the image contains an inst ance of the object, and, in case it does, to determine the transformat ion that relates the model to the image. The current approaches to thi s problem are divided into methods that use ''global'' properties of t he object (e.g., centroid and moments of inertia) and methods that use ''local'' properties of the object (e.g., corners and line segments). Global properties are sensitive to occlusion and, specifically, to se lf occlusion. Local properties are difficult to locate reliably, and t heir matching involves intensive computation. We present a novel metho d for recognition that uses region information. In our approach the mo del and the image are divided into regions. Given a match between subs ets of regions (without any explicit correspondence between different pieces of the regions) the alignment transformation is computed. The m ethod applies to planar objects under similarity, affine, and projecti ve transformations and to projections of 3-D objects undergoing affine and projective transformations. The new approach combines many of the advantages of the previous two approaches, while avoiding some of the ir pitfalls. Like the global methods, our approach makes use of region information that reflects the true shape of the object. But like loca l methods, our approach can handle occlusion.