Pedestrian detection based on Visconti2 7502

Abstract

Pedestrians are the important participants in traffic. pedestrian detection system is used to monitor the road condition of pedestrians in real time and provide warning information to prevent potential accidents, and is the significant component of the Advanced Driver Assistance Systems. How to detect pedestrians in real time and accurately is a hot topic. Aiming at the that the traditional research mainly stays in the research of algorithm and the poor real-time performance, or high cost. In this paper, a driver assistance system based on Toshiba Visconti2 series vehicle image recognition processor is proposed, which uses monocular camera to capture image, CoHOG to extract pedestrian features and SVM classification for real-time pedestrian detection, TBD for pedestrian tracking. The experimental results show that the system can effectively detect the pedestrian and meet the real-time requirement, the whole system is low cost and portable to use in vehicle environment absolutely.