Research On Complex Background Lane Line Detection Based On Fractional Theory And Least Square Method

With the rapid development of transportation industry,the lane line detection technology based on road images has become a hot research topic.In the process of the lane line detection,it might encounter all kinds of complicated situations,then a road image will contain some complex background,such as low brightness,fuzzy edges,oil pollution,etc.In these cases,the existing lane line detection method cannot meet the demand of the road detection,it is hard to detect the lane lines.According to the above problems,this paper describes the three major technologies,such as image denoising,lane line extraction and linear fitting.The main tasks are as follows:(1)In view of the characteristics of low road image brightness and fuzzy edges,this dissertation uses the algorithm based on fractional integration to denoise the lane line images.Not only is the noise of the image removed,and is image edge information retained,but also will it not make the image fuzzy.(2)The lane line extraction algorithm which based on maximum inter-class cross entropy and improved adaptive fractional order differential for the images with low brightness and fuzzy edge is proposed.By analyzing the characteristics of an image,the optimal threshold of image is got by use of the maximum inter-class cross entropy method,then the average gradient in the field of eight pixels with the optimal threshold is compared to select adaptive order of the fractional differential operator to extract the lane lines.(3)This paper uses the improved least square method to fit the lane lines,and within the scope of vision a lane line is defined as a linear model,the fitted line uses the vanishing point to connect the lane line,and the Kalman filter is applied to predict the lane area to reduce the search range,to improve the instantaneity of the detection.Because the proposed algorithm is generally applied to low contrast images and fuzzy images.Therefore,the paper selected two representative images for the line detection,and the results are compared with that by the algorithms in references,the experiment results show the effectiveness of the algorithm and its advantages.