Lane Marking Detection And Lane Positioning Technology Based On Machine Vision

Vehicle lane-level positioning refers to the positioning technology that uses the lane as a basic unit to determine the location of the road on which the vehicle is located and its specific lane.It is the key and foundation of advanced driver assistance systems,driverless technologies,and intelligent transportation systems,as well as automotive related services such as car-hailing,car sharing,and social car networking.However,at present,there are relatively few vehicle lane-level positioning devices in the market,usually only equipped with high-end vehicles,and they are often expensive,complicated in equipment,have large changes to vehicles,are difficult to transplant,and have poor generality.Therefore,this thesis is devoted to providing a low-cost,simple-to-use,and versatile vehicle lane-level positioning service for vehicles.In order to achieve the above-mentioned goals,a vehicle lane-level positioning method based on smart phones and machine vision is proposed in this thesis,which makes full use of the advantages of wide range and strong functions of smart phones,as well as the low cost,convenient installation and maintenance,and versatility of machine vision.In this method,the smart phone is used as a comprehensive perception computing platform.Firstly,the road-level positions of the vehicle are acquired via the smartphone.Then the road video image is collected through the camera,meanwhile,lane markings are detected from it.Next,according to the lane markings,a detection of vehicle lateral offset is performed to obtain the lane positions of the vehicle.Finally,road-level positioning results and lane positions are merged to obtain the lane-level position of the vehicle.Through these steps,vehicle lane-level positioning is achieved.The key to this method is to determine the position of the lane where the vehicle is in the road,whose premise is the accurate detection of lane markings.Therefore,the detection of lane markings is discussed in detail firstly in this thesis.First,according to the different lane markings and road conditions,vehicle driving scenes are divided into highway scenes and city road scenes.Then,under the highway scenes,the highway lane marking detection(HLMD)algorithm based on ROI selection,graying,filtering,binaryzation,edge detection and Hough transform is proposed.Next,under the city road scenes,for problems such as uneven lighting,shadows,blocking,road damage,and obstacles,the city road lane marking detection(CRLMD)algorithm is proposed,based on the HLMD.In CRLMD,the binarization process is improved,road area detection and filtering in the HSI color space,morphological processing,polar angle and distance constraints are also introduced.Finally,the two algorithms are tested and verified on the actual road,which show high accuracy and robustness to the environment.The thesis is followed by the design and implementation of lane positioning algorithms.First,the movement of vehicles in the road is decomposed into longitudinal movement and lateral movement.Then,according to the fact that lane change is only related to vehicle lateral movement,the lateral motion of the vehicle is analyzed in detail.Next,the concepts of lane part,pressing state and pressing period are introduced to distinguish between lateral motion in lane and lateral motion between single lanes.Finally,a method of lane positioning based on initial lane position and vehicle lateral motion detection is proposed.By updating the lane position at the time when the pressing state is terminated,the robustness of lane positioning is greatly improved.Finally,the thesis gives a brief introduction to the development of lane-level positioning APP based on the Android system.And then,the APP was tested on the main roads in Xi'an City,in an afternoon with fine weather and normal traffic conditions.The test results show that the APP can determine the road position of the vehicle at present and the specific lane in the road,that is,the APP can realize vehicle lane-level positioning,the algorithm performances good and the scheme is successfully implemented.