99 Pedestrian Detection Based on Leg Feature Available to Purchase

Nowadays the research on Road Safety is receiving more and more attention, and the pedestrian detection is the significant part of such research. Our study focus on the feature of pedestrian's legs because they have an obvious difference from the background when people walking across the road. In paper, we present a novel idea that contributes to low-false alarms and high detection rate. First, we detect the pedestrian based on Haar cascade classifier which is trained by the leg samples. Second, we verify the pedestrian through the HOG descriptors. The results from experiments show our method gets a detection rate of about 87% with few false alarms.