Based On The Characteristics Of The Traffic Sign Recognition Applied And Research

Intelligent transportation systems through the sharing of information between people and cars, cars and trucks, cars and road vehicles with sensors and other equipment, making people more harmonious relationship between the three of the vehicles, roads, and therefore countries have startedand to develop rapidly. As an important part of the Intelligent Transportation Systems, traffic sign recognition system has also been given due attention and development, the system can help the driver to determine the road traffic conditions, so as to effectively increase traffic safety, so, its research has important theoretical significance and practical value.This article is based on image processing technology for traffic sign detection and recognition technology research:(1) Based on the color information of the traffic signs, traffic signs in natural conditions to convert the image to RGB, HSV and SVF color model space for color segmentation. Finally,through the comparison and analysis segmentation results of three model. The results show that, based on the color information of the segmentation can remove to dry most scrambling background, while supported by a candidate area containing the target region.(2) Use binarization segmentation algorithm based on color segmentation to further separate the background of the image and the candidate target area, then de-noising and smoothing the Binary image. Using mathematical morphology filled candidate region and complete candidate region and exclude most of the non-target candidate area by the area threshold; gets the bounding rectangle based on regional connectivity and extract the traffic signs in natural scenes. Analysis of shape attributes of our existing common traffic signs, shape attributes as a basis for classification of traffic signs.(3)Analyzed the feature selection and extraction theory, the Hu moment and zernike moment, result show them with a rotating, scaling, translation invariance, so we can selection them as Classification feature of traffic signs image recognition, and then the extracted by calculating the traffic Hu sample signs unchanged Sandy, and Zernike unchanged Sandy characteristic values and characteristic data normalization processing. In view of the traffic signs belong to the small sample size problem, statistically-based VC dimension theory and design principles, based on SVM is designed classifier, and the classifier parameters (C,g) the optimization process, in order to get the optimal classification results. Finally, the design-based classifier to complete the training and prediction model to be classified logo recognition experiments, and the experimental results are analyzed.Based on previous research and analysis, design algorithm for block diagrams, using modular structure design of traffic sign recognition systems and corresponding interfaces and experimental verification of the feasibility of the system.