A multi-feature fusion based traffic light recognition algorithm for intelligent vehicles

Traffic light recognition is a key technology for intelligent vehicles and Advanced Driver Assistance Systems (ADAS). This paper proposes a multi-feature fusion based real-time traffic light recognition algorithm for intelligent vehicles. In the region of interest determined by the vanishing line, technologies including color segmentation, blob detection, and structural feature extraction are employed individually to obtain a set of candidate locations. A fusion algorithm is developed to integrate these results and compute a score for all these possible locations of traffic lights. The score of each candidate denotes its probability of being a traffic light. The final detection is achieved by fusing its score with temporal and geographic information. Extensive experimental results on a real intelligent vehicle show that the proposed algorithm is effective and efficient.

1 Introduction

Researches on intelligent vehicles are becoming deeper and more extensive in recent years. Lots of efforts have been put into developing techniques such as lane detection, obstacle detection and vehicle motion planning. To drive smoothly and automatically in urban traffic scenes, the intelligent vehicles must obey traffic rules which demand a real-time and robust traffic signal recognition system.