Research On Lane Detection System Based On Fisheye Surround View

With the continuous development of society in recent years,the number of cars has increased,and traffic accidents have also followed,so the safety of driving has also received increasing attention.Among the many traffic accidents,most of them are caused by drivers’ inattention,speeding or blind spots.Therefore,it is of great significance to solve the problem of blind sight and lane line detection.In this study,based on the car 360°surround view system,the forward-looking fisheye camera is reused for lane line detection,and related technologies and algorithms are studied in detail,the main research contents are as follows:(1)Because the extraction of the corner points of the image in the calibration toolbox of Zhang Zhengyou’s calibration method needs to be performed manually,the accuracy of the corner points and the extraction efficiency are not high.Therefore,this paper proposes an improved algorithm for automatically extracting corners based on SUSAN to improve the speed and accuracy of detecting corners in a checkerboard.(2)Of the road information captured by the fisheye front camera of the 360 °surround view system,since the lane line occupies a small proportion,the smaller the proportion,the lower the accuracy of information acquisition when detecting information.Therefore,this paper proposes a lane line region extraction method that divides the area of interest into blocks,thereby improving the accuracy of detecting lane line information.(3)In view of the low computational efficiency of the Hough transform and the susceptibility to interference,which cause lane line misdetection and missed detection,based on the segmented straight line model,on the basis of limiting polar angle,the parameter space of polar angle and polar diameter is used Block processing method to improve the calculation efficiency and accuracy of Hough transform.The experimental results show that the lane-line detection technology based on fisheye surround can be fully integrated with panoramic vision and lane-line detection technology,and the improved and optimized algorithm proposed in this paper canimprove the accuracy of the algorithm,have intuitive visual effects and obtain Good experimental results.