A Visual Words Selection Strategy for Pedestrian Detection and Analysis of the Feature Points Distribution

An effective and efficient visual word selection method based on Bag-of-features (BoF), which can be applied to the pedestrian detection problem, is proposed in this paper. We first calculate the difference in the total appearance frequency of each visual word in pedestrian and non-pedestrian images. Visual words that exhibit greater absolute values are more efficient for pedestrian detection, and are thus selected. The effectiveness of the proposed method is validated by analyzing the distribution of selected feature points. Through this analysis, we find that discriminative feature points for pedestrian images are mainly located about the lower body, whereas those for non-pedestrian images are mainly located in background areas. Experimental results show that, using the proposed method, the detection rate for the Daimler-DB datasets exceeds 92.5%, whereas the miss rate is less than 6.8%. More-over, the time required for learning and detection can be reduced by approximately 50%, with no significant degradation in precision, using the proposed method, even if only 40% of the visual words are selected.