|
As science advances, the development of urbanization, the number of vehicles and travel a significant increase in the number of road traffic safety issues become increasingly prominent. In this context, people began to put intelligent transport systems (Intelligent Transportation Systems, referred ITS) studies. Intelligent transportation system is the one including image processing, digital signal processing, electronic technology, artificial intelligence, information technology, pattern recognition, communications technologies and systems engineering, integrated system, it protect the safe operation of transport system has very important significance. Road traffic sign recognition (Traffic Sign Recognition, referred TSR) system as an important vehicle subsystem intelligent, semi-automatic and automatic vehicle has become an important part. Its role is in the process of moving vehicles on road traffic sign image acquisition and recognition, in a timely manner to make instructions or warnings to the driver, guide the driver to properly control the vehicle in order to maintain smooth traffic and prevent accidents. So research road traffic sign recognition system has important theoretical significance and practical value. Select this speed limit signs as the research object recognition algorithm is mainly based on the reliability and validity. The main contents of this paper include: (a) the reasons due to environmental factors, such as weather conditions, traffic information, it will cause the image to obtain color distortion, size is too big or too small, blurry picture. In this case, it must be pretreated. In this paper, pretreatment included in the HSI space I channel histogram equalization algorithm, so the picture's brightness enhancement; using image scaling technology to adjust the image size; through the image blurred image restoration technique that becomes clear (2) According to this study limits speed characteristic color of the mark, we use the RGB color space and the HSI color space, the color hue H channel segmentation, the background and the separation of the target area, generating the binary image. In the RGB color space of the target area can be split up significantly, and spend less time deficiency is noise points; while in HSI space H channel segmentation, because black is no color, so it can be very red split Well, very little noise points, but the black part is totally ignored, and because there is a space from RGB to HSI space conversion process, so relative to the RGB color space segmentation consuming. Through comparative analysis, this paper uses the RGB color space partitioning. Since the presence of noise, this paper introduced after color segmentation median filter to remove outliers noise. (3) According to this paper, the speed limit signs circular features, this paper roundness parameter extraction based on the image of a circular target area. Degree of circularity of each region is calculated based on the area and perimeter calculation obtained. The method computes a small amount, real good, and high reliability. After extraction of the shape, using the projection method to remove a circular outline, and then the rest of the target area on mathematical morphology fill the edges with the edge of the hole and remove the burrs, so that the target area smooth. (4) Identification of the content of this work is the most important part. The first target feature extraction through the template, and then were used template matching and BP neural network to identify the target. By comparison, BP neural network recognition accuracy is better than template matching.
|