Research On Lane Detection Technology Based On Improved RANSAC Algorithm

In recent years,with the development of science and technology and the continuous progress of society,the number of automobiles has shown an upward trend year by year.While automobiles bring great convenience to people's lives,the driver's misoperation has led to more and more serious road traffic accidents and irreparable losses to people's life and property safety.In order to improve the active safety performance of vehicles in road driving and reduce the incidence of road traffic accidents,the research and development of intelligent vehicles with various active safety technologies and advanced assistant driving technologies has become a common research hotspot in academia and industry.Lane detection is an irreplaceable key technology in intelligent vehicle active safety technology,which has high research value and important significance.Therefore,how to accurately detect the lane area in the road has become a research hotspot in the field of intelligent vehicles.Based on the analysis of the advantages and disadvantages of existing traditional lane detection methods,this paper proposes a lane detection method which can accurately detect lanes in complex environments and based on improved Random Sample Consensus(RANSAC)algorithm.The main research contents are as follows:Firstly,this paper studies the corresponding transformation relationship between two-dimensional image and three-dimensional space.Combined with the internal and external parameters obtained by camera calibration,the bird's-eye view of road image eliminating the perspective effect of image is obtained by inverse perspective transformation calculation.On this basis,the Region of Interest(ROI)of lane line under inverse perspective transformation is obtained by using the method of region of interest selection based on the actual physical width information of lane.Then,a method of lane line effective feature point extraction based on ridge detection and adaptive threshold is proposed.Compared with traditional edge detection operator,the ridge detector constructed in this paper can accurately detect lane line central feature points in complex road environment with interference,and effectively eliminate interference noise according to the adaptive ridge threshold.Then,the parabolic lane model is selected based on the analysis of various Lane models,and the lane model is fitted by the improved RANSAC algorithm.This paper improves the traditional RANSAC algorithm from three aspects:(1)The feature data set is partitioned by the method based on histogram statistics and the location constraints of feature points.Optimize the selection;(2)optimize the sampling process by using the dynamic feature distance-guided sub-area sampling method;(3)optimize the criteria for judging the effective Lane model.Finally,compared with the traditional RANSAC algorithm,the three improvements minimize the impact of interference noise area on Lane detection,and greatly increase the accuracy of model fitting in complex road scenes.In this paper,a large number of experiments are carried out to verify the proposed method in complex road scenes such as uneven illumination conditions and the presence of shadows,road signs and vehicle interference.The results show that the lane detection method based on improved RANSAC algorithm has good robustness and accuracy.