Research On Lane Detection Technology For Structured Road With Monocular Vision

This article carries out a research focused on lane detection based on monocular vision and be aimed to propose a robust lane detection algorithm that can be applied under many complicated road traffic environment s.This article introduces the existing research status of lane detection.Based on the state-of-the-art algorithms on lane detection,we propose a lane detection that synergizes multiple visual cues with the a-priori knowledge and spatial-temporal information.The main contents of the thesis are as follows:1.This article begins with the contour feature detected in bird's-eye view with a fast line segment detector.Compared to the contour detection method based on contour detection operator,the line segment detector can achieve sub-pixel accuracy,and acquire the lane marking features under low contrast,uneven illumination and far view field,which contributes to improve the lane detection range.2.In terms of the lane detection,to improve the robustness of the lane detection under complicated road traffic environment,we conduct the Classification-Generation-Growth-based operator to the detected lines by synergizing multiple visual cues of the lane markings,such as the gradient,length and location information.If the outspread linear result satisfies the threshold requirement,it will be the final detected result of the linear lane.3.In terms of the fitting model,the thesis uses linear-parabolic model wit h dynamic joint point to match the curved lane,which is suitable for variable width lanes and stochastic lane shapes.4.In order to enhance the tracking response performance of the Kalman filter,the thesis designs two types of noise models to improve the tracking robustness and response speed under different road traffic environment.Through the compare to the other lane detection algorithms and the test of road images under different scenes,the experimental results show that the proposed lane detection can exhibit the robustness under complicated road traffic environment and satisfy the demand of practical application of lane detection.