Research On Lane Line Detection Algorithm Based On Monocular Vision

Lane line detection based on machine vision is an important application in driverless perception.According to the geometric features of lane line,other curve models such as straight line,parabola and hyperbola can be adopted,and lane line can also be detected according to its color features such as brightness and width.Due to the change of light environment and the complex lane environment caused by pedestrians or vehicles blocking the road surface,the diversity of lane line shape is caused.Therefore,the design of a lane line detection algorithm with high identification accuracy has become a research hotspot of vehicle assisted driving system and unmanned driving system.On the basis of the existing lane line detection algorithm and image algorithm,this paper proposes the lane line recognition algorithm of point-by-point detection in view of the fact that the current lane line recognition methods mostly adopt the global lane line model,which cannot adapt to the situation of the lane line in the complex environment.(1)in terms of the primary information extraction of lane lines,the existing preprocessing methods are compared and analyzed,and an image grayscale processing method based on the longitudinal and transverse characteristics of lane lines is proposed,which is used to deal with the problem of poor robustness of the ordinary lane binary segmentation method in the environment of variable illumination.Experimental results show that this method can extract more lane line information in complex environment and eliminate redundant noise.(2)due to the perspective effect of monocular camera,the lane lines photographed have different widths in the transverse direction and tend to converge in the longitudinal direction,which is not consistent with the characteristics of lane lines in the real environment.Therefore,the aerial view of lane line is obtained on the basis of camera calibration,which is beneficial to lane line identification.(3)in order to make the lane line continuous testing results,on the basis of binarization processing,aspirant line distance is calculated pixels and edge points recently distance on the X axis direction,this method can not only make the continuous part continuous,more can give the approximate location of a lane lane line and the halfway point,provide the particle filter with operating conditions.(4)in terms of lane line identification,this study aims to improve lane line detection in the urban environment.A two-dimensional lane line model with smooth lane width was established,and the parameters of the two-dimensional lane line model were predicted and updated by particle filter on the basis of the row distance,so as to achieve the purpose of lane line identification.This method can accurately detect lane lines in urban road environment with more noise interference.The off-line video data collected in this paper are validated by the algorithm,and the road scenes in the expressway and urban environment are tested.The experimental results show the robustness and effectiveness of the algorithm in this paper.