Algorithm Study for Pedestrian Detection Based on Monocular Vision

For intelligent vehicle and driving assistant system, pedestrian detection technology is an important research field to avoid dangerous traffic accidents. This article puts forward a pedestrian detection algorithm based on edge symmetry. First, lane recognition method is used to get area of interest (AOI) ahead of vehicle. Second, as pedestrian legs have prominent vertical edge symmetry, their symmetrical axis can be acquired through vertical edge extraction in the AOI. Combined with pedestrian transcendental features, the candidate pedestrian could be locked. Third, the candidate pedestrian will be validated based on the gray symmetry and local entropy. The experiment results show that the algorithm is effective, reliable and robust.