Chord angle representation for shape matching under occlusion

A shape description method based on chord angle representation was proposed to solve the problem of shape matching under partial occlusion, meanwhile balancing retrieval accuracy and computational efficiency. A chord angle descriptor was defined based on the angle between two chords for each sample point, which could be used to describe an open contour by its self-contained property. Then, a match cost matrix was constructed by computing the L1 distance between descriptors of all the sample points on two open contours. Finally, the similarity between two contours was obtained by the integral image algorithm and the partial shape matching result was achieved. The experimental results on MPEG-7 and Kimia216 shape databases indicate that this method is robust to the partial occlusion, and the computational efficiency and the retrieval accuracy are both essentially improved as compared with other partially occluded shape matching algorithms. The retrieval accuracy of proposed partial contour matching method reaches to 83.63% and increased by 19.09%. It concludes that this proposed method meets the requirements of shape matching and object recognition in efficiency, accuracy and ability of anti-occlusion. ©, 2015, Chinese Academy of Sciences. All right reserved.