Lane And Road Marking Detection Algorithm Based On Computer Vision

In recent years,with the development of computer technology,the research of Advanced Driving Assistant System(ADAS)and Autonomous Driving Systems have attracted extensive attention from scholars at home and abroad.Lane and road marking detection is the most basic and important part of these systems.Lanes are line graphics used to divide the drivable area for vehicles.Road traffic signs are arrows,texts,patterns,etc.painted on the road.Some computer vision algorithms can be used in the task of lane and road marking detection,which plays a key role in supervising of driving behaviors such as lane keeping,lane changing and turning.This thesis has carried out research and improvement of lane and road marking detection algorithm based on computer vision.The road information is acquired by the camera,and the lanes and road traffic signs are respectively detected by improved image processing algorithms or deep learning algorithms.The work content of this paper mainly includes the following points: First,I collected driving videos,and construct the road traffic sign data set through screening and labeling.Second,I implemented the function of lane detection based on the improved image processing algorithm,and propose a lane detection tracking process according to the rules of the lane appearing in the video.I have improved the detection speed by setting the traversal rule of pixel points.Third,I explored the feasibility of the deep learning-based target detection algorithm using in the road traffic sign detection task,and improved the Mask R-CNN algorithm.I have solved multi-scale changes and deformations of targets by introducing feature pyramid networks and deformable convolution.Fourth,I proposed two detection and tracking strategies for road traffic signs detection task based on video,and improved the speed of the system by introducing KCF tracking algorithm.In order to analyze the performance of the proposed algorithm and strategy,I designed multiple sets of contrast experiments.In addition,some experimental result graphs under various road conditions are shown to demonstrate detection performance of the algorithm.Finally,the algorithm and strategy proposed in this paper are proved useful.Practicality,the algorithm and strategy can well realize the function of lane and road marking detection.