A TECHNIQUE FOR PEDESTRIAN DETECTION BASED ON MOTION FEATURES

Abstract

Pedestrian detection is an important issue in many application areas of image processing, such as traffic monitoring, intrusion detection, self-driving car... In this paper, we present a pedestrian detection technique based on extended Haar features combined with weak classifiers are implemented based on the Adaboost algorithm to make decisions. These features have been calculated based on the difference between pairs of images over time. The technique has been implemented and demonstrates the effectiveness on the 2001 PETS database.