Study, Based On The Multi-feature Images Of Traffic Signs Detection

As economy is developing faster and faster nowadays, the problems it brings have become tougher and tougher. The heavy traffic, which is thought to be the greatest, has been concerned about world widely, for they are restricting the modernization and urbanization. Aiming at solving this problem, researches of ITS (Intelligent Transportation System) are in progress and importance has been attached to its key part, i.e. RSR (Road Sign Recognition).This paper discusses problems from three perspectives, namely road image preprocessing, road sign detection and the classification of road signs in natural scene.Algorithms using the differences between components in RGB color model and threshold in HSV color model are stressed in this paper and the option advice of algorithms for color segmentation is also introduced in nature scene. As the following, the best method to detect the edge is raised by tests of edge detecting and the detection of road signs based on shape feature is analyzed. On account of the round road signs, detection algorithm based on improved Hough Transform method is introduced while template matching algorithm related to symmetry and algorithm based on fuzzy rule are presented aiming to triangle traffic signs. Based on the analysis and researches of the color and shape features of road signs, segmentation and retrieval approaches in RSR are raised and it turns out to be a better way compared with detecting using single feature. And at last, based on the feature extraction and the SVM method, the third part mainly focuses on the three-layer recognition and classification of road signs.Visual C++6.0 is applied in this paper to analyze the validity of the above algorithm. It shows that great recognition ability can be gained from the algorithm and the outcomes are satisfactory. Due to the shortcomings in the process and the algorithm, further improvement is needed for better researches.