Research On Lane Detection Algorithm

With the rapid development of artificial intelligence,unmanned driving,as an important application field,has made rapid progress gradually.As the leading companies,Baidu and Google have made outstanding achievements in this area,which has greatly promoted the development of intelligent transportation and other related industries.In order to better assist driving,providing more secure and reliable algorithm mechanism is the current development goal of related industries.Lane detection,as the early work of intelligent driving,plays an irreplaceable role in the fields of automatic parking,anti-collision warning and intelligent driving.Research on lane detection technology emerges in endlessly from different organizations and companies.For lane detection technology,it will eventually be applied to the part of automatic driving,not only to detect road conditions,which needs further research to achieve.In this paper,the traditional image processing method is used to detect the lane line under monocular vision.The parallel coordinate system based lane line detection method is innovatively adopted.And the linear classification method based on connectivity analysis is proposed to detect the lane line,which has great progress in the real-time and accuracy of the algorithm.The main contents of this paper are lane line preprocessing algorithm,lane line detection process design and lane line model.(1)Preprocessing algorithm.The image preprocessing methods used in this paper include image segmentation,color space conversion,image enhancement and edge detection.Each step of the preprocessing method can accele rate the efficiency of the algorithm to a certain extent.Image segmentation technology can remove some interference information,reduce the search range of subsequent processing,;color space conversion is mainly to unify the lanes of different colors in the road,using the mapping of RGB color space to HSV color space to transform yellow lanes into white lanes;image enhancement algorithm can protrude the edge information in the image.A better adaptive detection algorithm can be obtained.The Canny operator is used for edge detection,and the Canny operator uses double threshold to close the edge of the broken lane to form a loop.(2)Lane line detection process design.The more difficult problem in lane line detection is the distinction of line curves,which is faster and occupies the majority of lane images,so it is an effective method to distinguish line curves.In this paper,the lane line is divided into straight line and curve by the method of connectivity analysis,and different models are used to detect the classified image.In addition,the continuity between frames can be used as the basis of lane line stability and lane tracking,and can further enhance the accuracy of lane line.(3)Lane line model detection.On the basis of lane line classification,lane line is classified into straight line and curve,which are detected separately to adopt different models.In the case of keeping the speed of the linear detection part,a better curve model is adopted to further increase the accuracy of curve detection.In this paper,the lane line detection method based on parallel coordinate system is used to replace the traditional Hough line detection algorithm,which saves more time and speeds up the efficiency of the algorithm.For the curve part,in order to better fit the lane line model,the cubic B-spline model is selected.This algorithm can better adapt to the curvature of lane line and improve the accuracy of detection.