Traffic Signal Recognition Using On-board Color Camera for Safety Driving Assistance

Abstract

This paper proposes a reliable vision system to detect traffic signals and to recognize their status. Its goals are to recognize the traffic signals robustly at any hour, and to achieve a detection range of more than 100m. It detects active signal lights mainly by utilizing color information to work during both day and night. Various cues, such as the shape-based signal detection and knowledge about the illumination order of the traffic signals, are integrated to distinguish false positives from the detected candidates. Quantitative evaluation using image sequences captured on urban roads has proved the effectiveness of the proposed system.