3D matching using statistically significant groupings

Vision programming is deened as the task of constructing explicit object models to be used in object recognition. These object models specify the features to be used in recognizing the object as well as the exact order in which they have to be used. For 3D recognition, in the absence of grouping information, the number of bases (model feature/image feature correspondences) that must be examined before a match is found is prohibitively large. By exploiting the relationships between features, we can avoid having to consider a potentially large number of bases. The automatic programming framework 5] helps us in ordering model features based on their utilities such as detectability and error rate that are derived from training data. Examining model features in the order speciied by this framework leads to minimal numbers of bases being considered before a match is found. In this article, we describe a vision programming approach to matching 3D models to 2D images. Our system considers feature clusters instead of individual features and dynamically orders unmatched feature clusters based on the existing state of the match. The dynamic feature cluster ordering is achieved through the use of a new dynamic cost function. The automatic vision programming framework is general enough to be used by any feature-based recognition system, and in this article, it is shown to lead to dramatic improvements in the performance of a correspondence-based object recognition system 14].