Design And Implementation Of Real-time Lane Detection System

With the rapid development of artificial intelligence,especially several well-known companies have done a lot of research on unmanned vehicles.For example,Google and Baidu make a great progress in the field of unmanned aerial vehicles,which have stimulated the current development of intelligent transportation industry greatly.Providing a safer and more reliable driving strategy ha s becoming a common purpose of all walks of life.As a basic work for unmanned vehicles,lane detection technology plays an important role in automatic parking,anti-collision warning and automatic driving,so the research of lane detection technology is also increasing.However,the ultimate goal of lane detection technology is to be a part of the automatic navigation system,rather than the traditional lane change instruction,of course it will take longer to improve the technology.In this paper,the lane detection technology is studied from the perspective of image processing,and a new lane detection algorithm based on maximum straight line is proposed,which has a great improve in real-time and accuracy.The general content of this paper is divided into three parts: preprocessing,the maximum straight line detection algorithm,and the lane reconstruction.? Preprocessing.In this paper,the preprocessing includes image segmentation,color space conversion,image enhancement,median filtering and edge detection.In order to improve the real-time performance of lane detection system,the image segmentation is as far as possible to remove some of the interference information,even if the removal of part of the effective information,the maximum straight line algorithm still play a role.Color space conversion is mainly to take into account the existence of yellow lane,but the traditional algorithms are based on the study of white lane,in this paper,the use of RGB to HSV color space conversion can effectively turn the yellow lane into white lane.Image enhancement is used to enhance the high brightness lane area in the image.Median filtering is used to remove the isolated bright spots in the image after image enhancement,while effectively preserving the effective edge of the image.Edge detection using the current best canny operator,which can connect the broken edges based on the use of dual threshold.? The maximum straight line detection algorithm.The maximum straight line detection algorithm include the maximum straight line detection algorithm,the small area matching tracking algorithm,the weighted update algorithm and the lane changing algorithm.The maximum straight line detection algorithm based on the maximum straight line to search for the innermost line of image,which passed the gray verification is judged to be correct,otherwise,there is no lane line in this frame picture.As a navigation algorithm,the weighted update algorithm is used to update the lane with the spatial and temporal characteristics of the lane,preventing excessive jitter and mutation.Lane changing algorithm is used to detect lane change,and change lane tip is given when the car is changing lane,in this paper,using the distance difference between the left and right lanes to judge the lane changing or not.? Lane reconstruction.Lane reconstruction includes lane stability algorithm,lane tracking algorithm,curve fitting algorithm and reconstruction evaluation.Lane stability algorithm mainly aims at the lane drift caused by the weighted update algorithm,which can correct the deviation lane line in a certain range and improve the accuracy of lane detection.The small area matching tracking algorithm can match the maximum straight line detection algorithm well,searching for the most matching straight line near the line of the previous frame,and setting the jumping mark to deal with the lane change.Curv e fitting algorithm is mainly aimed at the actual road conditions in the curve,the use of linear model to fit the curve is obviously wrong,so this paper uses the straight line plus Bezier curve model to fit the curve.Reconstruction evaluation is used to evaluate the gap between the detected lane line and the actual lane line.However,it is difficult to evaluate for the virtual lane.Therefore,the method of manual labeling is used to judge the correctness of lane line detection result.