Research On ADAS Lane Detection System Based On Video

The current status of traffic conditions in China are large quantity of accidents and high mortality.Both of ADAS(Advanced Driver Assistance System)and intelligent driving technology could solve the traffic safety problems.But they have not been popularized and applied widely at present.Because some functions are still at the research stage,and domestic research status lags behind foreign countries'.Lane departure detection system is an indispensable and essential part in ADAS and intelligent driving.So it is necessary to study it.This article aims to excogitate a lane detection system which could draw lanes in real time and realize lane departure warning(LDW).Firstly,the ADAS,intelligent driving and lane detection system are analyzed,then the lane detection system and the problems being faced are analyzed too in this paper.Secondly,the image preprocessing,edge detection,lane drawing,lane departure warning function module are studied.In the module of image preprocessing,using the weighted average method with specific coefficients to perform image grayscale.Next,enhance image by means of Gauss filtering and histogram equalization in smoothing filtering.In the module of edge detecting,analyze the technical principle of edge detecting at first.Then adopting improved Sobel operator to do the edge detecting after making a comparative analysis of Roberts operator,Prewitt operatror,Sobel operator,Laplacian operator and Canny operator.In the module of lane drawing,propose a method which could find vanishing points and interesting regions at first.Then put forward a concept of candidate points and representative points,as well as the lane line estimation method.In the module of lane departure warning,propose a mid-line departure warning algorithm after making a comparative analysis of several existing models of departure warning algorithm,such as CCP,FOD,TLC,EDF and so on.Then the experiment is carried out and the result is analyzed.Next,for the implementation of system's hardware and software,the hardware is mainly described as a whole.In the software,mainly adopts organization software structure which composed of three modules,such as underlying driver layer,lane detection middleware layer and lane detection application layer.All of them are contacted each other through API interface.Each of them has a low level of coupling.Finally,the system is tested.The experimental platform is introduced,and the test data set is analyzed.Then,the functional integrity,accuracy,stability and various road conditions are tested.The test results show that the system can achieve the function of lane drawing and departure warning.Meanwhile,the accuracy is high and the stability is good.