Natural Scene Detection And Recognition Of Traffic Signs,

With the rapid development of China's economy, cars have been widely spread into the thousands of families. When cars give people convenience, we also have a clear understanding that the traffic accidents take great harm to some families. Therefore, it is necessary to study intelligent transportation technology to help drivers reduce accidents.Automatic detection and recognition of traffic signs as an important part of intelligent transportation, get more and more attention. In order to improve the natural scene of traffic signs of detection and recognition rate and precision, the paper studies the following:(1) Traffic sign detection. In the natural conditions, the traffic sign images has more interference, In order to improve the detection speed, firstly, this paper use the color characteristics of traffic sign to segment the image, and then use the contour tracking technology to eliminate the small noise, Finally, using the traffic signs of circular and rectangular shape characteristic to locate traffic signs. This paper also introduces the visual attention model to obtain the image visual attention focus, and then given traffic signs graph, using similarity distance that based on Top-Down model to locate signs;(2) Extract traffic signs invariant moment. After detected signs, this paper use the improved invariants moment of HU and Zernike moments to extract the invariant moment of the traffic signs;(3) Identification of traffic signs. In traffic sign recognition, the paper uses images identification based on Hausdorff distance and fuzzy C Mean to identify traffic signs. Experiments show that fuzzy C Mean has better recognition effect than the image matching, and also has better value for application.The above algorithms are realized on the platform VC++combination of opencv. Experimental image proved the effectiveness and feasibility of the algorithms, the traffic signs are rapidly detected and identified.