Research On Algorithm For Traffic Light Detection And Recognition

Abstract:With the development of the society, the traffic management gradually becomes intelligent. And the traffic device is also on the way of intelligent era. The appearance of unmanned vehicles on the road and intelligent driver assistance will promote the progress of intelligent transportation system. As traffic lights are an important part of traffic signs, its intelligent detection and recognition play an invaluable role. This thesis focuses on the traffic lights detection and recognition, thus designing a system with better robustness and real-time. The main work of this thesis is as follows:1. Put forward the algorithm of cascade match to improve the progress of traffic lights’detection. The proposed algorithm detects traffic lights and its back area respectively in different color spaces, and locates traffic lights through cascade match. This algorithm can solve the problem of blocking traffic lights device.2. Extract the characteristics of traffic lights by improving2DPCA algorithm and combing with Gabor wavelet transformation. Through adoption of modular2DPCA algorithm based on within-class weight average and comparison with ordinary2DPCA algorithm in experiments, this thesis comes to a conclusion that the algorithm has better performance with stronger identification capability and lower time complexity.3. Put into practice the traffic lights detection and recognition algorithm. This thesis proposes a traffic lights recognition system which consists of input module, detection module, recognition module, display module and communication module. The experiment shows that the proposed system has a high recognition rate for traffic lights and better robustness and real-time.