Hough Clustering Technique for Surface Matching

2) Roof edges which indicate a discontinuity in surface normal but not in depth. Roof edges could be further classified This paper presents a 3-D multiple object reco~nition as either convex or COncave. technique using Hough clustering. The objects are modeled as polyhedra and the input image is a range image containing instances of the modeled objects. The features used for matching are surface discontinuity features namely dihedral junctions. Qualitative reasoning based on qualitative attributes assigned to scene features is shown to be an important aspect of the Hough clustering algorithm presented in this paper. The resulting algorithm is robust and well suited for multiple-object scenes with partial occlusion.