Recognition by matching dense, oriented edge pixels

This paper describes techniques to perform efficient and accurate recognition in difficult domains by matching dense, oriented edge pixels. We model three-dimensional objects as the set of two-dimensional views of the object. Translation, rotation, and scaling of the views are allowed to approximate full three-dimensional motion. A modified Hausdorff measure is used to determine which transformations of each object model are reported as matches. The use of dense, oriented edge pixels allows us to achieve a low rate of false positives. Techniques to prune the search space are used to obtain a system that is efficient in practice. We give results of the system recognizing object views in intensity and infrared images.