Curve Lane Detection And Departure Warning Based On Machine Vision

With the continuous increase of the number of vehicles,the traffic safety problem caused by the misoperation of drivers is becoming more and more serious.The intelligent vehicle technology represented by the intelligent driver assistance system can reduce the traffic accident caused by the driver's mishandling effectively.This paper proposed a curve lane detection and departure warning algorithm based on machine vision.It can detect the lane in front of the vehicle and calculate the lane equation in real time.When the vehicle departure from the current lane or trend to departure from the current lane,it can issue early warning signals to driver,reduce traffic accidents caused by driver's fatigue or distraction,and improve driving safety.In this paper,before the lane departure warning,the lane equation is estimated by lane detection.Firstly,in the image preprocessing stage,the image is corrected for distortion,and then the corrected image is subjected to inverse perspective mapping(IPM)to obtain a top view image of the road ahead of the vehicle,and the noise of IPM image is eliminated by bilaterally filter.Then the Sobel operator is used to detect the lane edge in the IPM image,and the lane feature image is extracted based on the lane line width model and the edge detection image to eliminate the interference of non-lane information.After acquiring the lane feature image,the lane feature points were searched,and the least square method was used to fit the feature points and obtain the lane equation.In order to improve the accuracy of lane feature points extraction,the Kalman filter is introduced to track the feature points of the lane lines,establish the connection of feature points between the frames before and after,and use the lane line equation to establish the dynamic region of interest for subsequent images.Finally,the feature points of the lane centerline are calculated according to the lane feature points tracked by the Kalman filter and fitted to obtain the lane centerline equation under the pixel coordinates.At last,the proportion between the inverse perspective image and the actual coordinate space is determined,and the front road equation,the curvature of the road ahead are calculated.Different types of the lane departure warning models have been analyzed,then lane departure warning algorithms is formulated.The algorithm is implemented in Visual Studio.The video image is acquired from a monocular camera mounted on the test vehicle for testing.The test results show thatthe proposed lane detection algorithm can detect the lane marking lines accurately under different road conditions and show good real-time and accuracy.The extraction accuracy of lane centerline equation can meet the test requirements,the lane departure warning algorithm can perform lane departure warning to drivers accurately.