Research On Lane Line Detection And Tracking Method Based On Machine Vision

In recent years,with the development of highway construction and the popularization of automobiles,it has brought great convenience to mankind and created considerable economic benefits.However,the traffic safety problem is becoming more and more serious,and its impact can not be ignored.As an important guarantee of driving safety,the intelligent auxiliary driving system of the vehicle guides the driver to drive safely in the form of warnings or deviations from warnings to avoid traffic accidents.As an important part of the safety intelligent auxiliary system,lane line detection and tracking has important research value and significance.This paper mainly focuses on lane line detection and tracking for in-depth analysis and research.Firstly,the paper introduces and analyzes the research status of lane line detection and tracking algorithm at home and abroad.As the most obvious characteristic of roads,the paper studies the detection algorithm of straight lane and the detection method of curved lane respectively according to the characteristics of structured highways in China.In visual navigation,lane line detection requires good real-time performance and high accuracy of road positioning.Therefore,it is of great significance and value to add tracking algorithm to auxiliary driving system.By studying various algorithms of intelligent vehicle system at home and abroad,this paper focuses on the lane detection algorithm and tracking algorithm.In the lane line detection,in order to remove the lane line detection useless redundancy information,and enhance the accuracy in the lane line detection algorithm,need to be collected road image preprocessing,and every step of preprocessing algorithm simulation comparison,select out the best effect,suitable for this operator.In order to make the results more close to the real shape of the lane line,and consider the computational complexity of algorithm,straight lane line detection algorithm on inverse perspective transformation algorithm is used to improve the probability of Hough transform algorithm.Then,by comparing with the traditional Hough transform algorithm,it proves that the algorithm has strong applicability and good detection effect.In bending the lane line detection algorithm,this paper studies the three fitting methods,including CatMull exemplifies cd-rom spline curve method,the algorithm adopts curve model,can under the condition of the linear model is not enough joint corners to get better detection result and the simulation was carried out on the curve fitting,up to the bend lane detection results.In the lane tracking,three tracking algorithms are analyzed,and the detection algorithm with tracker is improved in real time and robustness.Through the initial frame detection lane line,record its information and predict the position of the frame.The system is constantly updated to identify areas of interest in dynamic changes.Then,the lane line is detected and fitted to achieve the tracking purpose.Because this article mainly studies the background of the environment is in high speed,such as city structured road,in the simulation,performance contrast,application of Kalman filter to track the lane line,can not only achieve tracking,and improved the accuracy of the test results.Finally,the proportion of image cropping is proposed in this paper,and the tracking algorithm is improved according to the cropped image,which improves the accuracy of tracking.