Research On Lane Detection In Complex Environment

In driving assistant system of intelligent vehicle,environment awareness is an important part,and it is also the core function of intelligent vehicle to obtain surrounding environment situation and its own driving state.In the process of improving the safety of automobile driving,most of environmental perception functions rely on visual information.Lane line detection technology,the preliminary work of intelligent driving,is essential for lane departure warning,lane keeping,path planning,etc.Therefore,the lane line detection based on machine vision has important research significance.The lane line information is extracted from road images collected by vision sensor,and the research is mainly aimed at lane line detection and tracking methods in complex environment.The main research contents of this paper are as follows:(1)Determine ROI,use Gaussian filtering to denoise white noise and random noise in road image,use the Canny operator to extract image edge information and perform straight line detection based on the progressive probabilistic Hough transform.The part that does not meet the slope range is eliminated to reduce interference.The region of interest in original image is determined by vanishing point detection based on the least square method,and the bird eye view containing lane line information is obtained through inverse perspective transformation,which improves subsequent image processing speed.(2)In the part of lane line detection algorithm,firstly,lane line features are extracted from bird eye view,and obtain the target feature binary image through the color vector filtering algorithm and HSL color space conversion.For other objects in the environment that are similar to the target color,such as direction signs,etc.the rectangle threshold is set to eliminate them,and perform image fusion processing based on region growth on the resulting binary image to obtain a feature point set.Comparing the feature extraction method with Otsu method,the results show that the proposed method has better anti-interference ability in complex scenes,and can extract more complete Lane features.After analyzing the lane line fitting method,the improved RANSAC algorithm is proposed for the shortcomings of the cubic polynomial fitting method,and combined with the Bezier curve for lane line fitting.For the problem that fitting effect of lane line with larger curvature deviates from actual lane line,the Bezier curve optimization algorithm is proposed.The detection results show that the optimization algorithm used to identify lane lines in complex environment is more accurate and has higher fitting accuracy.(3)For the lane line tracking algorithm,lane line tracking based on Kalman filter is used to solve the problem of missing or false detection in lane line detection process due to factors such as wear and occlusion of the lane line in a complex environment and uneven illumination.Due to good continuity of lane information in adjacent video frames,the detection information of adjacent effective frames can be used to complete the corresponding lane within a certain range,which makes the detection results more accurate.The Kalman filter is used to update region of interest in real time,which reduces the calculation of image processing and further improves real-time performance and accuracy of the algorithm.Aiming at the problem of lane detection in complex environment,this paper uses multiple data sets and video streams to verify the effectiveness of the detection and tracking method.Experimental data and results show that the proposed algorithm can detect lane effectively and has good robustness.