Research On The Lane Detection Method For Structure Road

With the increase of the number of cars in the society,traffic safety has become a serious problem.At present,it has become a hot subject to develop and study the assistant system of safe driving by means of sensor technology,pattern recognition,image processing and deep learning.Through the study of driving assistance technology,people have the chance to get access to the intelligent driving or even the automatic driving,so as to make sure thier safety and improve comfort when driving.Lane line detection is a key research subject of advanced driving assistance system.Correct identification of lane line can improve the road safety and reduce the hidden troubles.Based on the structured road image,this thesis proposes a structured lane recognition method.This thesis mainly focuses on two aspects: the first one is the extraction of the region of interest in road image;another one is the extraction of lane line.In order to improve the correct rate of lane recognition,researchers usually use a variety of methods to identify the lane area of interest.In this thesis,Gabor filter is used to detect the texture direction of the image to identify the structured road area.There are two sub subjects: the detection of road vanishing point and the extraction of road boundary line.Aiming at the voting algorithm which takes up a lot of time resources when detecting the vanishing point.this thesis proposes an improved voting method,which significantly reduces the time consumption and improves the real-time performance of the system.Aiming at the problem of road boundary line search,this paper roughly determines the location and direction of the road boundary line according to the statistical information of voting points,and adopts an iterative updating method,which effectively improves the detection accuracy and the real-time performance.For the problem of lane line extraction,the thesis followed the method of "feature extraction-model fitting".We firstly identify the road area,in which the region of interest of the inverse perspective transformation is selected,and the top-view of the road is obtained by the inverse perspective transformation.In the top-view image,the possible areas of lane line are located and the edge points of lane line are extracted as the input samples of fitting algorithm.For model fitting,this paper proposes a fast fitting method based on parabola on the basis of RANSAC method.Finally,based on the discussion of the above methods,the algorithm flow of lane line detection is given,and lane detection experiments are carried out on the image test set.The experimental results show that the algorithm is effective and robust.