Study On Road Detection Technology Based On Dynamic Image

The main two components of the road tracing including road track and obstacle detection, the actual structure of the road often can be divided into structured and unstructured. As the structural shape of the road rules, there have been many times in research, so this study aimed at unstructured roads. According to the characteristics of unstructured road such as irregular shape, no road line markers, and severe affection caused by lighting and water spots. In order to solve the above problems, this article from a road segmentation and boundary identification has been studied in these two-pronged approachs:First, the road region segmentation, thus bringing the road detection problem is transformed into the image segmentation. Image segmentation refers to the similarity in accordance with certain criteria for the image into several meaningful, non-overlapping areas. Threshold of calculation is simple and has been widely applied to image segmentation. Although the Otsu method in road detection experiments, demonstrated good performance, can effectively inhibit surface defects, gray uneven lighting, water-soaked on the impact test results, with good usability, robustness and real-time, but the segmentation result is not satisfactory, so the Otsu algorithm to improve, with the Otsu method to improve the image divided into several times, and then refer to the region by means of the road to consolidate information from the region belong to the road so as to enhance the accuracy of the road zoning.Second, through the identification of the road boundary detection problem is transformed to the road edge detection problem. First, the original dynamic image edge extraction, and then use Otsu method for binary image segmentation operator Canny edge extraction image filtering, the final use of Hough transform and least-squares method, the edge of the image by depth-first principle of curve fitting, extraction image number of the main road curves, the use of selected road refinement precise path of the border area. The results show that the original of dynamics image pre-processing, use the improved Otsu thresholding method, in the segmentation process of the introduction of the concept of fuzzy rough sets, and will split the value of the image after the two pairs of Canny operator. Extraction of the edge of the image is filtered to eliminate the complex background or light and shadow produced by some of the road edge, and finally used Hough transform and least squares curve fitting of the image. Compared with the original approach, using the improved algorithm processing images, not only improve the performance of the algorithm is adaptive, but also to make a complex environment, road recognition accuracy and robustness has been optimized to meet the actual application requirements.