Pedestrian detection by stereo vision

Abstract

This paper describes a human detection by stereo camera at road scene. Human detection is one of the important functions of driver support systems or autonomous driving cars. This technology uses sensors, for example, a stereo camera, LiDAR and so on. In this study, we develop a stereo camera system with two inexpensive web cameras. The system detects human candidates using stereo and verifies them using monocular image features. We also compare several stereo matching algorithms in terms of the accuracy and the calculation time.