Research On Lane Detection Algorithm In Intelligent Driving

Witih the continuous increase of the number of cars,it not only brings convenience to human life,but also leads to traffic accidents,traffic jams and other adverse consequences.For this reason,the intelligent transportation system emerges at the historic moment.The most important significance of intelligent transportation system is to realize unmanned driving.As one of the key technologies in the field of intelligent transportation,lane detection technology is of great significance to the realization of intelligent driving,lane departure warning system and vehicle anti-collision system.Based on the research and analysis of the current situation of lane detection technology at home and abroad,this paper explores and discusses the detection of lane.The following is a brief description of the main research content of this paper.Lane detection is easily interfered by non-lane information such as light change,trees,texts on the road surface,road background on both sides of the lane,as well as bad weather in rainy days.Firstly interested in image region selection and weighted average gray,smooth filtering processing,in dark and light the lane under the image grayscale average,points under the condition of dark and light image enhancement processing,through the image processing to highlight the edge of the lane line information,weaken the road lane area such as background,the words.For further access to the edge of the lane characteristic information,the classical edge of first order differential operator and the edge of second order differential operator made a detailed analysis,and the high and low double threshold for Canny edge detection algorithm is not selected,through the introduction of normal distribution with the classical Canny detection algorithm selection from the combination of adaptive threshold,the lane of the processed image edge detection,the simulation results show that the improved Canny algorithm can adaptively select double threshold,and compared with the traditional algorithm can better to keep lane edge information unless the lane information.In order to make the lane detection effect more accord with the requirements,the real-time is higher.In this paper,the shape of lane is determined according to the diversity of its shape,which is divided into straight line and curve part for detection.For the straight lane,inverse perspective transformation and improved Hough transform are adopted to fit the edge of the lane.Through simulation analysis,this algorithm is more practical and real-time than the classical Hough transform.In the curve detection part,the boundary points detected by the improved Canny edge are used to form the feature point set,and uses the least squares method combined with parabola,RANSAC combine three curve model to the simulation of curves respectively,the results show that three times of curve fitting based on RANSAC algorithm can accurately in the corner fitting at the same time,relative to the least squares fit for larger curvature curves fit has better effect.Finally,kalman filter algorithm is adopted to track the lane in the process of lane detection,because the lane may be blocked,wear and other problems such as missed detection and wrong detection.The research background of this paper is mainly highway,city structured and township semi-structured road.The tracking algorithm of kalman filter can not only achieve the tracking purpose,but also effectively improve the accuracy of lane detection.