Robust traffic sign detection in complex road environments
Bin Tian, Ran Chen, Yanjie Yao, Naiqiang Li · 2016
Traffic sign recognition is an important topic in driver assistant system and intelligent autonomous vehicles. Traffic sign detection is a critical step, whose performance greatly affect the performance and computation cost of traffic sign recognition. In this paper, we propose a traffic sign detection method based on a scoring SVM model. First, traffic sign color and color gradient are extracted according to their color characteristics. Then, the shape of traffic sign is computed by a voting scheme, yielding shape score maps. After that, the score maps of traffic signs are used to train a SVM model. Compared with single traffic sign score value, the score map is more efficient to verify the existence of a traffic sign. Finally, the trained SVM model is used to detect traffic signs. Experiments show that the proposed method is more effective than the voting based method to detect traffic signs.