Traffic Signs Detection And Recognition Based On Signal Decomposition

With social progress and quick development of economy, the numbers of vehicles in a lot of center cities exploded recently, which also inevitably leads to some other problems like traffic jam, driving safety gets more and more attention, being a part of Intelligent Traffic System(ITS). There are quite a number of related computer vision techniques in ITS, traffic sign detection and recognition is one of the most important techniques.The goal of traffic sign detection and recognition is during driving, the system can detect the potential traffic signs and obtain messages the traffic signs carry correctly. This field has become an important research subject because of its’ high practical value-improving driving safety. After reading related papers and being inspired by algorithms of other fields, the innovations of this paper include:(1)improving the adaption of detection stage, the system can detect traffic signs in most situation. (2)during recognition based on sparse representation, a new connected dictionary is trained to avoid disadvantages of multilevel classification problem. (3) recognition based on non-negative matrix factorization(NMF) is utilized to improve physical meanings of matrix.Photos taken in a moving car are used in detection stage. To compare the performances with popular algorithms in this field, GTSRB signs library, which contains more than 40 classes traffic signs, is used to train and test. The experiment results shows that the algorithms our paper proposes balance the real time and rate of recognition, and is robust to illumination and rotation.