Research On Lane Departure Warning Algorithm Based On Monocular Vision

With the increase of car ownership,more and more people pay attention to the safety problem of automobile.The emergence of automotive advanced driving assistance systems provides solutions to automotive safety issues.Based on the low cost and high integration principle,the paper integrates the lane detection system with the lane departure warning system based on the monocular vision sensor.A lane departure warning system that integrates the lane line detection function and the departure warning decision function is designed to better realize the early warning when the car deviates and achieve the purpose of assisting driving.First,camera calibration is performed and then video frames are captured through the car driving video stream to meet the human eye’s refresh frequency.Lane detection and visual output are performed after image preprocessing for each frame.At the same time,the state prediction is carried out through the pixel-level analysis of the extracted position information of lane pixel,combined with Kalman filter.Finally,the improved lane departure warning algorithm based on the driving habit statistical model and dynamic TLC algorithm is used to make a hierarchical warning decision.The main contents are as follows:(1)An improved fusion algorithm for the edge detection of lane lines is proposed to optimize the preprocessing process.The improved edge detection algorithm based on lane line connectivity characteristics can better highlight the feature of strong pixel connectivity of lane lines.Better-preprocessed images can be provided as input for lane line detection.(2)An improved lane line detection algorithm based on the dual sliding window mechanism is proposed by studying the lane line detection algorithm and the characteristics of the lane line in the image.The algorithm uses a window mechanism to search for lane lines and improves the window sliding method.It uses an overlapping sliding mechanism to search for lane lines and makes full use of pixel information.(3)Considering the different driver’s perceptions of dangerous situations caused by various driving habits,this paper proposes a TLC lane departure-warning algorithm based on probability and statistics model.The algorithm introduces a statistical probability model and a hierarchical early warning mechanism to improve the deviation early warning decision model.The vehicle status information predicted by the Kalman filter is used to budget for the time of departure from the line through the TLC algorithm.Besides,according to the performance of the estimated deviation from the pressure line in the deviation early warning decision-making model based on the statistical probability model and the grading early warning mechanism,effective grading early warning is carried out.Experimental results show that some related algorithms in the improved lane departure warning technology proposed in this paper have a certain degree of performance improvement compared with the previous algorithms.An improved fusion algorithm for lane line edge detection is proposed to better express lane lines’ edge characteristics.An improved lane line detection algorithm based on the dual sliding window mechanism is proposed to improve the utilization of image information,and to a certain extent,avoid the misdetection caused by the information loss problem.A TLC lane departure early warning algorithm based on probability and statistics model is proposed,which can realize the hierarchical early warning of different driving habits.The above-improved algorithm performs well in lane line detection and deviation warning scenarios,and effectively improves the algorithm’s accuracy.