Abstract:
In this paper we propose a novel heuristic block-generating algorithm for reducing redundant features and increasing detection efficiency in pedestrian detection.Firstly,the detection window is divided into overlapped blocks with variant sizes.Then these blocks are sorted according to the stability of the histogram of oriented gradient for each block.Finally,HOG features from the most stable blocks can be extracted.Experimental results show that the proposed algorithm is faster and more accurate than the existing ones.