Selection of Bins on Histograms of Oriented Gradient

Abstract

In this paper, a method has been proposed, which focused on the problem of pedestrian detection in static images, adopting the Histogram of Oriented Gradient features proposed by Dalal and Triggs. A novel features set based on selecting oriented bins, which is the key-point to the Histogram of Oriented Gradient's performance, is presented. Due to the comparison of three different processing bins, we achieve a proper testing result at 86.63% by SBS algorithm. The simulation results shows that the detection rate of proposed method outperforms slightly than the original HOG performance, yet decreases the running time of the whole procedure.