A two-stage strategy to introduce spectral matching into recognition of occluded objects

When recognizing partially visible objects in a scene, a good global decision should be made based on locally gathered features for their recognition, since global information is corrupted. This local to global nature of occlusion recognition leads us to spectral matching technique. Unfortunately, spectral matching algorithms are not desirable for noisy data set from cluttered scene. In this paper, a top-down procedure is introduced into spectral matching for the recognition of occluded objects. Based on the two-stage strategy, both appearance and geometric information are taken into consideration. It is shown that the improvement has been made for spectral algorithms to recognize occluded objects.