Research On Prohibition Traffic Sign Recognition Method In Intelligent Driver Assistance System

Driver Assistance System is one of the key technologies to solve traffic safety issues. The Prohibition Traffic Sign Recognition, a capital component of the DAS system, plays an important role of forewarning drivers the requirements when driving and assisting safe driving. So it has got more researchers’attention in the field of Intelligent Driver Assistance System in these years.Color segmentation, shape detection and kernel pattern recognition are the key steps in the process of Prohibition Traffic Sign Recognition. But there are some disadvantages in the existing methods, those are the range of threshold value for red is inaccurate, too many pseudo-circle in circle detection and excessive dimension of classification feature in recognition. Some methods are proposed in this paper for improving the disadvantages and the details as follows:(1) Prohibition Sign Color Segmentation. According to the characteristic of the prohibition sign is that the basic color is red, some measures are taken as follows. Firstly, the red pixels are randomly sampled from prohibition sign in different weather conditions, and then the discrete point set is mapped to RGB, HSV, HSI, YCbCr and Lab color space, secondly, a novel evaluation criterion for the dispersion degree of all red pixels in those color spaces are presented. RGB and YCbCr are preferred as the optimal and suboptimal color space after evaluating the dispersion degree and transition time. Finally, the threshold value of red pixel in those two color spaces is redefined through statistical analysis and the most appropriate elliptic envelope method, and then the results demonstrate the effectiveness of the proposed threshold.(2) Multiple Circular Signs Detection. Firstly, the best method of image pre-processing is adopted by comparing the methods in reprocessing phase. Secondly, the connected areas are filtered by the geometric constraints of circle. The boundary information of the persisted block is extracted as an approximate radius, and then a circular template is dynamically constructed by the radius. Finally, the potential center location is confirmed based on the coordinates of the maximum in correlation matrix, and then the multi-prohibition traffic sign detection in complex scenes is achieved through template matching method.(3) Speed Limit Signs Kernel Number Recognition. First, this feature extraction of digital on speed limit sign kernel is achieved by improving the existing handwritten digit character feature extraction methods, meanwhile, in order to get the most simple matching template, the dimension of the extracted characteristic attribute set is reduced by the rough set attribute reduction and value reduction method. Finally, the fast recognition is achieved through the template matching algorithm, and the results demonstrate the accuracy of the method.The method proposed in this thesis obtains better performance in detecting and locating multiple prohibition traffic sign in complex scenes, and it can extract more stable characteristics and the rate of decline of the characteristics reach to87.7%. Consequently, it realized the fast recognition of speed limit sign efficiently.