Road Traffic Signs Detection And Recognition Technology Research In Driving Assistant System

With the development of the society, the number of motor vehicles is greatly increasing, traffic jams,traffic accidents,transportation efficiencies are becoming more and more serious.Intelligent Transportation System (ITS) is considered to an important meas to solve the problems, which has great advantages and vast market value, at the same time as an important branch of ITS—Traffic sign recognition system has also get corresponding development, This paper has done the key research on traffic signs detection and recognition algorithms.First, this thesis puts forward a traffic signs detecting method which integrating of the color and edge characteristics.First this method according to the characteristics of traffic signs in the video images, using SVF color vector filter combining HSI color space to achieve the initial traffic signs segmentation, and then extracting the edge information of the single frame image, effectivly integrate the results of color segmentation and edge image and substantially eliminate the interference of background, finally, the paper utilizes morphological method to process the integrating iamge and gets a more complete target edge region, all of these lay the foundation for subsequent traffic signs positioning.Second, by analyzing the characteristic of traffic signs shape, this thesis puts forward an improving algorithm.For the round ban and direction traffic signs, a four directions template matching algorithm based on symmetry features is proposed to overcome the problems of traffic signs detecting difficulty when existence blocks; for the rectangle direction signs and triangle warning signs, starting from the corner features and combining with their own geometric structure characteristics to complete traffic signs detecting, also modified the searching strategy, and experiments show that this method is effective.Finally, analysing the human visual goal recognition process, combining with human visual sensory characteristics proposed a hierarchical recognition mathod based on significant features. The general procedure and specific method of hierarchical recognition are introduced, on this basis determining the significant of property characteristics by analyzing the BP network convergent situation which is coursed by different properties, and then determine the significant degree of target attribute characteristics, at last they are applied to traffic sign hierarchical recognition in the video images.Experiment shows that the hierarchical recognition method not only can achieve better recognition results, but also meet with the general process of human understanding things.