| With the development of society and economy and the popularity of cars, traffic accidents occur frequently, the vast majority of which a,re due to human factors, and most of those were related to ignore traffic signs for drivers. As an important part of the traffic signs, the speed limit signs mainly provide speed limit information, which has a strong reminding ability on driving behavior. As far as be concerned shape, speed limit signs mainly contain rectangle speed limit sign and circular speed limit sign. Rectangle speed limit sign are used in U.S. Circular speed limit sign are used in Europe, Asia and Australia. The effective recognition of the speed limit sign can remind drivers in time to the change of conditions and allowing speed for driving in front of the car to help drivers control the vehicle effectively. So it can improve the driving safety, and reduce traffic accidents as far as possible. Setting speed limit sign is a way to control the speed of car for the world of countries. But, owing to a variety of reasons, during the processing of driving, it is difficult that driver does not always find the speed limit and content timely and accurately. Therefore, it is the automatic detection and recognition of speed limit sign that is of great significance for safe driving. In this paper, we describe a traffic sign recognition system which is adapted for rectangular U.S. speed limit signs. However, it can be used in circular speed limit sign detection and recognition.As being in complex outdoor environmental conditions, speed limit sign’s detection and recognition are often vulnerable to the weather, light, tilt, fade and similar to the background, furthermore, the performance of the algorithms need to consider of not only recognition rate but also real-time. So it is quite difficult to design the algorithm for speed limit sign detection and recognition. By summing up a large of research methods at home and abroad, this thesis has improved the detection and recognition algorithm, and has designed and implemented a new type of speed limit sign detection and recognition algorithm.In the phrase of detection, to avoid of light and the change of weather influence on the image color information, the stroke width transform is presented to detect every digit region inside of speed limit sign from input image. In order to get more edge details, and remove the false edge and texture edge, it is put forward that the edge detection operator based on the surrounding suppression. The proposed algorithm can overcome the influence of many factors, such as light; solve the similar backgrounds and obstacles. Accuracy and reliability for the detection can be improved.In the phrase of segment and recognition, all digit candidates will be recognized, In order to improve the reliability of detection and accuracy of recognition, the mechanism based on the extreme value is put forward. And it can accurately find boarder of digit. Then1-a-1SVM classifiers are used to recognize the target.In the integration of digits and re-searching, if there is another digit in a digit’s neighboring, they will be merged according to the rule, otherwise, we will re-search in its neighboring, if a digit is found, they will be merged too; otherwise, it will be deleted.In the phrase of multi-frame fusion, when rectangle speed limit sign appear in the video sequences, it will continue during a period. After the process of multi-frame fusion, it can increase the recognition rate and reduce the false recognition rate.Experimental results shown that the method had a high detection and recognition rate with rectangle speed limit signs in car video and certain robustness with complicated scenes, including lighting, weather changes and tilt, and was a simple procedure with runs near real-time. So this thesis found a practical solution to solve the existing problems and difficulties of the rectangle speed limit sign recognition, and laid a good foundation for the following research on traffic signs recognition. |
PDF Full Text Request