Road Traffic Sign Detection And Recognition Technology Research

Traffic Sign Recognition (TSR) is an important component of Intelligent Transportation System (ITS). Because of the complexity of road traffic, the research on TSR is not mature. It also need in-depth study and discussion to develop the efficient and practical TSR.In the 1980s, the research on TSR had already begun abroad, but in China the field of a late start. In the increasingly well-developed road traffic system, TSR will have a wider range of applications. Therefore, the researches on TSR are very important and have a practical value.According to the inherent color information and shape characteristics of traffic signs, the research of this paper which is to detect and identify traffic signs includes the following four:1. By comparison and analysis, the paper implemented the traffic sign image preprocessing algorithms. By median filtering, histogram equalization and double Gamma correction algorithm to solve the problem of color distortion from the impact of weather conditions and the external environment in the image acquisition process.2. Depth study of traffic signs detection algorithm based on color information. Firstly, for the traffic signs which have been pretreatment image, main research is the conversion of color information and the extraction algorithm in HIS color space and RGB color space based on the analysis of traffic signs'color information; Secondly, it gives OTSU and segmentation results based on the fixed threshold of histogram analysis; Finally, it use mathematical morphology for image denoising and achieve traffic signs detection.3. This paper proposed a method for eliminating the false traffic signs based on shape characteristics. For the geometry characteristics of traffic signs, by obtaining the enclosing rectangle in the binary image connected region, the symmetry analysis for each connected region is used to determine the target area and the precise positioning of traffic signs have been completed.4. Design and implemented TSR based on SVM. This paper used to meet the requirements of the rotation, translation and scale invariance feature vectors-Hu invariant moments and Zernike invariant moments as the match features, and design SVM classifier recognize traffic signs.This article has programmed to complete these algorithms using the Visual C++6.0, and the effectiveness and feasibility of these algorithms were compared and analyzed. Experiments show that:Algorithms for complex scenes in various outdoor conditions are very good effects in detection and identification, with good recognition reliability and lower rates of false detection and missed.