Recognition of partially occluded objects using maximum curvature points

Partially occluded objects are recognized from a 2-D image through the use of maximum curvature points on the image boundary. The vertices of high curvature on an occluded object are classified by the objects which are hypothesized to be involved in the occlusion. A heuristic method is developed for computational speed. Two typical examples are given to illustrate the accuracy as well as the simplicity of the heuristic method.