Robust Lane Detection And Tracking For Structured Road

With the continuous advancement and development of society,people's material and spiritual life has gradually improved.Cars have become more and more necessary transportation for people to travel,but while bringing convenience to people,more and more traffic accidents have caused People's personal safety and property security are threatened.Lane detection is a key technology for Driver Assistance Systems.It provides information on the position of the vehicle and the lane position,and thus determines the area in which the vehicle travels.Therefore,the accuracy,real-time and robustness of lane recognition algorithms are the key issues to be solved in the practical application process,which is of great significance.The research topic of this paper comes from the national key research and development project “Development and Industrialization of Intelligent Assisted Driving Technology for Electric Vehicles(2016YFB0101102)”.In the actual road environment,the road scene is complex and changeable.Facing the interference of different factors on the lane line,the main research is how to pass the visual inspection.The tracking algorithm solves the problem of identification of lane lanes and satisfies good accuracy,real-time and robustness.The main work done in this paper through a large number of experiments is as follows:(1)In the image preprocessing module,an optimal algorithm flow is given: inverse perspective point transformation,grayscale,OTSU binarization and morphological denoising,effectively eliminating some non-lane line interference.Provides more efficient input for subsequent clustering algorithms.(2)In the lane line detection section,clusters based on density clustering are divided into different clusters according to the color and geometric features of the lane lines.Then,an improved RANSAC algorithm is proposed to complete the lane line extraction based on the parabola model.Aiming at different environmental disturbances,an optimization strategy was developed to realize the effective detection of lane lines in complex environments.The effective detection of lane lines under working conditions ensures robustness and effectively improves the accuracy and real-time of lane line detection.(3)In the lane tracking part,comparing different tracking algorithms,through a lot of experiments and theoretical analysis,finally,Kalman filtering is selected to track the lane line model to ensure the stability of the system.Through the prediction and update of the lane line model parameters,the effects of missed detection,false reduction and illumination changes are effectively overcome,and the accuracy and robustness of lane recognition are further improved.Finally,based on Visual Studio 2013 and Open CV 2.4.9 development software,this paper realizes lane line detection and tracking under different working conditions,such as illumination changes,lane line discontinuity,lane marking and vehicle interference,and vehicle lane change scenarios.The results show that the algorithm has good accuracy,realtime and robustness.