An Effective and Robust Pedestrians Detecting Algorithm & Symposia

In this paper, we present a pedestrian detection approach using spatial histograms of oriented gradients feature. As spatial histograms of oriented gradients consist of marginal distributions of an image over local and global patches, they can preserve shape and contour of a pedestrian simultaneously. There are two main contributions in this paper. First of all, we expand the histograms of oriented gradients features from single-size to variable-size which can capture local and global feature of pedestrian automatically. We call theses feature as the "spatial histograms of oriented gradients". Secondly, we employ histogram similarity and Fisher criterion to measure discriminability of features and select some discriminative features to identify the pedestrian. SVM classifier is constructed to train the selected features from target and surrounding background. The proposed algorithm is tested on some public database. Experimental results show that the proposed approach is efficient and rapid in pedestrian detection.