Recognition of occluded polyhedra from range images

Occlusion remains a major hindrance for automatic recognition of 3-D object s. In this paper, we address the occlusion problem in the context of polyhe dral object recognition from range data. A novel approach is presented for object recognition based on sound occlusion-guided reasoning for feature di stortion analysis and perceptual organization. This type of reasoning enabl es us to maximize the amount of information extracted from the scene data, thus leading to robust and efficient recognition. The proposed approach is based on a multi-stage matching process, which attempts to recognize scene objects according to their order in the occlusion hierarchy (i.e., an objec t is recognized before those that are occluded by it). Such a strategy help s in resolving some occlusion-induced ambiguities in feature distortion ana lysis. Furthermore, it leads to verification of object/pose hypotheses with greater confidence. Matching is based on a hypothesize-cluster-and-verify approach. Hypotheses are generated using an occlusion-tolerant composite fe ature, a fork, which is a pair of nan-parallel edges that belong to the sam e surface. Generated hypotheses are then clustered and verified using a rob ust pixel-based technique. Indexing is performed using distortion-adaptive bounds on a rich set of viewpoint invariant fork attributes, for high selec tivity even in the presence of heavy occlusion. Performance of the system i s demonstrated using complex multi-object scenes. (C) 2000 Pattern Recognit ion Society. Published by Elsevier Science Ltd. All rights reserved.