The Research Of Lane Detection Algorithm In Structure Load Image

With the rapid development of social economy and highway transportation system, the total number of cars is increasing and traffic safety problem is also very prominent. According to the traffic accident statistics, the main reason of lane departure traffic accident is inattention and fatigue driving of drivers. So it is necessary to design vehicle safety assistant driving system to warning drivers timely in case of deviation from the lane or collision to reduce injuries. The core of vehicle safety assistant driving system is lane detection and recognition system. Because of the complexity of the road conditions and the instantaneity of vehicle safety assistant driving system, this article focus on improving real-time and robustness of the lane detection system. We have done a thorough research about how to extract the road lane line quickly and efficiently in different environment based on morphological characteristics of structured road image.First of all, the background of this project is introduced and the frame of this subject is set up by analyzing the research status at home and abroad.Then, the road image collected by a CCD industrial camera is pre-processed. Those pre-processing steps including image graying, image denoising, image enhancement, image binarization, edge repair and related technology introduced in detail. A suitable pre-processing algorithm for this topic is made after comparing and analyzing common methods.In the end, the characteristics of lane line area are extracted. Natural dividing line between road area and sky background are used to select "seed" by vertical binary scanning.Area threshold is introduced to filter lane line in image, thus the lane line feature area is obtained. Detection and recognition of lane line are achieved by using inherent features of lane line. Directed simply connected chain is introduced by combining Hough transform with least squares method to merge effective run-length(in the same direction and belong to one lane line) and match all the lane include solid line and dotted line that improving the detection rate of lane line especially the dotted line.Experimental results show that this algorithm is effective in a variety of different conditions of road images and has a good robustness and the real-time.