Research And Implementation Of Real-time Traffic Sign Recognition System Based On Vision

With the growth of Chinese economy and the development of automotive, the quantity of private vehicles increases quickly in recent years. While enjoying the convenience, the traffic accident problems, because of speeding, has become more prominent day by day. As an important component of the Intelligent Transportation System, Traffic Sign Recognition System has become the development trend in academia. The function of system is recognizing the traffic sign, based on road scene images which is shot by the camera in the front of the vehicle and the theory of Computer Vision and Machine Learning. The purpose is to provide timely environmental information and the necessary safety warnings, and prevent accidents from happening.Based on summarizing and improving various methods, we have designed and implemented a real-time traffic sign recognition system. It is composed by two major parts: detection module and classification module.In the detection stage, we use the fast radial symmetry method to detect circle. The algorithm can overcome the impact of light, not like the method based on color. Also, the system will be real-time.In the classification stage, we use the method of an Improved Multi-class Support Vector Machines Based on Binary-tree, which can solve the problem of not-be-classified region, also it is more efficient. But there are some problems, for example, it has the effect of error accumulation and the structure of binary is the key factor to keep efficient. To solve these problems, we use the method based distance between sample to build the binary-tree, which can solve these problems.To solve the problem of complex computing of the method of Machine Learning, we selected feature based on F-score, which is not only suitable to Multi-class problem, but also can solve these problems. It can enhance the system-real-time effectively.At last, we selected a lot of road scenes and use the system platform of PC and ASB-S700to test the system. The experimental result shows it is a real-time and robust system.