Design and Implementation of Traffic-Scene Understanding System for Autonomous Vehicle

Abstract

In this paper, vision based scene understanding algorithm is designed for autonomous driving system. Integrated algorithm is composed of lane detection and lane type/position recognition part, road marking recognition part, traffic sign recognition and traffic light recognition part. Visual recognition result is delivered to vehicle control part for generating steering angle and speed value. Through the field test in the autonomous driving platform, we found that proposed system shows robust performance in real road and traffic environment.