Research On Lane Detection And Tracking Technology Based On Line Segment Detector

In recent years,with the development of automobile safety protection assistance driving system,the key lane detection technology of this system has gradually become the research hotspot of scholars.In order to overcome various interference effects that will occur in the actual scenarios,this study adds the lane tracking steps on the basis of lane detection,and makes full use of the spatiotemporal consistency of the road image,effectively improves the robustness and real-time performance of the whole algorithm.This thesis mainly studies the lane detection and tracking method under the structured road scene based on vision technique,the mainly work is as follows:First,determine the region of interest based on the camera mounting position and image size,excluding the large number of extraneous parts that exist in the acquisition road image.Based on the characteristics of only white and yellow lane line,a contrast enhanced weighted average grayscale method is proposed,and then the lane line is separated from the background part by using image processing technology such as image smoothing,image enhancement and threshold value.Second,a more robust and fast line segment detector(LSD)is used for lane boundary detection,and its advantages are verified by comparative analysis.Then,based on the characteristics of the line segments,an adaptive slope filtering method is proposed,and the screening effect is more reliable than the methods based on the vanishing point or the fixed slope range.Third,based on the characteristics of the color,length and quantity of lane line,a three-level overlapped classifier is designed to add the practical function of lane type discrimination.Considering the interference situation in the actual scene,the lane tracking technology is added to improve the robustness and real-time performance of the algorithm,and the dynamic region of interest is generated,hence the target detection area is further accurately positioned to reduce the processed size.Finally,through the test of the actual road image containing various situations,and through the comparison with other algorithms,the validity of the proposed algorithm is verified in this paper.