High-Speed Traffic Sign Recognition System On Mobile Devices

Car has gradually become the main means of transport for people to travel, but with the increasing number of cars, the incidence of traffic accidents is also increasing. In order to solve the problem of traffic safety, many countries have already started to research Intelligent Transportation Systems (ITS) for assisting traffic safety by high-tech. Within ITS, traffic sign recognition (TSR) is one of the most important parts and it has important significance of study. On the other hand, with the arrival of the mobile Internet era, the mobile device has become a necessity. And Android operating system for mobile phones and tablet computers is one of the most popular systems. Therefore, this paper hopes to find both the accuracy and timeliness of the best traffic sign detection and recognition algorithm by comparing algorithms, optimize the Traffic sign recognition system partially. Finally the system will be built on Android platform to realize the TSR system for mobile devices.In traffic sign detection, this paper uses the feature that a sign having different color will have a specific shape correspondingly. As a result, firstly, pictures are filtered by red, blue and yellow in HSV space to produce three black and white binary pictures, after morphological processing, area and aspect ratio filter, and then are detected by the law of roundness and the law of rectangle corresponding to the color detection result. In traffic sign recognition, the paper studies PCA and HOG two feature extraction algorithms, and proposes to use PCA to reduce the dimensionality after using HOG to extract features, and use SVM to classify the traffic signs. Finally, the paper determines to use HOG+PCA method which has both accuracy and timeliness after experiments in the image library built by myself which contains 38 kinds of traffic signs. Meanwhile, the paper first proposes to use lane line detection to locate the road and then reduce the detection area according to the law that traffic signs mostly appear on the right side and top of the road. Following these, the TSR System is built on the Android platform.After tests about the TSR System on Android in natural scene, we find that the results of both accuracy of real-time test are very good.