Segmentation Algorithm Natural Scene Of Traffic Signs

With the rapid development of economic and society, the number of cars is increasing year by year. Especially in large cities, the increase in traffic accidents and environmental pollution intensifies almost become a common problem faced by our cities, it even becomes the burden of economic development. In this context, intelligent transportation technology has come into being and gets rapidly development. Traffic signs are important ancillary equipment for traffic safety. It has played an irreplaceable role in guiding pedestrians, standardizing transport vehicles and guiding route. Traffic sign recognition system is part of the intelligent transportation technology, which can be used in intelligent driver assistance systems and unmanned vehicles. Traffic sign segmentation is the important foundation and prerequisite of traffic sign recognition, and the result directly affect the recognition rate. So the traffic sign segmentation has become a key step in the traffic sign recognition system.In this paper, we study traffic sign segmentation methods from the aspects of color and shape, which using traffic signs images captured in the real scene as the research object. Traffic signs are segmented effectively by using both color feature and shape feature. The main research content and innovation are as follows:(1) Segmentation methods based on color feature which include RGB color space, HSI color space and OHTA space are given. Further, SVM classifier is used in the RGB space for improvement. Most of the non-traffic sign areas can be removed by color feature.(2) Images segmented by color feature are converted to gray level images. A multi-threshold segmentation method based on gray-scale histogram is proposed, by which the binary images are obtained. Then the morphological operations are performed on binary images.(3) According to the shape information of traffic signs, the contours information in the target image are extracted with chain code. On the other hand, the improved edge detection method in gray-level traffic sign images is used to extract the edge information.(4) The boundary invariant moments has been proved to be invariant to rotation, translation and scaling, so it is suitable for traffic sign image processing. The boundary value of invariant moments is used to determine the area of traffic signs.The experimental results demonstrate the algorithms effectiveness:firstly segmenting images in the color space; then getting the binary image by using adaptive threshold selection algorithm in gray level image; then using standard traffic signs contour information to create the shape template; finally determining the target region shape.