Object Detection Algorithm Based on Multiple Shape Templates

摘要

A new object detection algorithm based on multiple shape templates is proposed to overcome the effect of edge discontinuity and distortion in edge image. Horizontal and vertical direction templates are generated by using a one dimensional and a two dimensional filter in order, and shape distribution template is generated based on edge image. The distances between three pairs of templates are calculated to measure the similarity between the model and the region in testing image. The region in testing image with minimum distance to the model is regarded as the position of the object. Experimental result on ETHZ dataset shows that the proposed algorithm achieves higher detection rate than the other algorithms.