Eigenwindow method updated by a mean eigenwindow

This paper proposes an object recognition method where background and occlusion problems are analyzed without subtracting them from the object portion. Conventional eigenwindow method is updated by a mean eigenwindow in this paper for recognizing partially data-loss objects due to occlusion and/or destruction. In the proposed method, various similar poses including disturbed shapes of an object are stored in a particular window of an eigenspace referred to as the ´eigenwindow´ and, finally, mean of similar poses of each window is taken in order to obtain a generalized eigenwindow called the ´mean eigenwindow´. Therefore, a set of mean eigenwindows finally makes an eigenspace with respect to their pose variations. This mean eigenwindow is further used for recognizing an unfamiliar pose, that may also include partially occluded and/or destroyed shapes, and the object type itself. We have applied the proposed approach to various data-loss environments and the method has successfully performed the recognition of an object with up to 20% of occlusion and/or destruction.