Research On Road Traffic Sign Detection And Recognition Algorithm

With the rapid development of economy and the progress of science and technology, the car ownership has increased substantially in recent years, which also brings tremendous potential demands to develop smart Driver Assistance Systems. As an important branch of the Driver Assistance System, the research on traffic sign detection and recognition algorithms has received much more attention. However, due to the complexity of the real traffic scenes, we still face many obstacles to be tackled.In this thesis, we mainly focus our research on traffic sign detection and recognition within complex traffic scene, and the achieved research results are as follows:1. To deal with the traffic detection in complex traffic surroundings such as illumination variation, occlusion, and color fading problems, a traffic sign detection algorithm based on SVDD model was proposed. An adaptive discrimination based on the statistical histogram of S component in HSV space is first carried out to rapidly find the candidate traffic sign region. Meanwhile, the SVDD data description method was applied in HV color space to build multiple one-class discrimination models. Thus, different categories of traffic signs can be robustly detected. In addition, by the further shape analysis on the detected sign area, the false alarm rate can be drastically decreased, which leads to a better performance.2. On the basis of the above traffic sign detection, we proposed a traffic sign recognition method based on core content area matching. To boost the performance of traffic sign recognition, a sub-area containing the core content of traffic sign located in the detected traffic sign area was first extracted by using an adaptive binarization method. With the extracted core content area, the Hu Invariant Moments and SURF descriptor are used to fulfill the sign recognition by template matching between the core content area and traffic sing dataset. The experimental results validate the effectiveness of the proposed algorithm.3. Based on the development platform of VS2010, a robust and real time prototype system for road traffic sign detection and recognition is built based on the proposed algorithms.