Research On Autonomous Navigation Method Of Intelligent Vehicle Based On Machine Vision

The advancement of science and technology and the gradual upgrading of intelligent products have driven the continuous development of society.The research of intelligent vehicle visual navigation technology has been greatly promoted by the emergence and maturity of machine vision technology.This article focuses on the research and development of intelligent vehicle navigation technology for campus inspections,getting rid of traditional remote controllers,and making them capable of autonomous navigation.The key of visual navigation technology is to extract the lane information through the camera,correctly identify the lane line by using the algorithms,and calculate the navigation parameters according to the lane line to realize the autonomous navigation of intelligent vehicle.The main research contents of this paper are as follows.First,the Robot Operating System under Ubuntu is outlined,and the forward and reverse solutions of the two-wheeled differential intelligent vehicle are analyzed.The sensor configuration of the intelligent vehicle platform is given,the camera imaging principle is analyzed,the camera calibration is completed by Zhang Zhengyou calibration method,the camera's internal parameters are obtained,and the overall scheme of autonomous navigation of the intelligent vehicle is presented.Secondly,an improvement is made to the lane line detection algorithm based on probabilistic Hough transform.According to the actual installation angle of the camera,the area of interest was extracted,reducing the processing of redundant information;Due to the particularity of lane line color,the gray-scale transformation is used for the image by weighted average method;The bilateral filtering method is used to remove the noise points generated in the image acquisition;For areas with uneven illumination,an image enhancement method based on gray stretching is used to highlight the edge information of lane lines;For the edge detection process,Canny operator with strong ability to suppress noise and good edge information retention is selected.Because the probabilistic Hough transform method is not ideal for detecting the real-time and accuracy of lane lines,according to the lane model,four-point constraints are proposed from the perspective of lane line angle,length,and width.The line segment is fitted to extract the correct lane line.Experiments show that the improved algorithm based on this paper has higher accuracy and better real-time performance.Then,the principle and common problems of artificial potential field method are discussed,and the following improvement measures are made based on the problems.In a dynamic environment,according to the relative speed of obstacles and intelligent vehicle,a speed repulsive force field is introduced,which increases the reliability of obstacle avoidance and solves the problem of unsatisfactory dynamic obstacle avoidance;Since the traditional distance factor has the defect that the turning radius of the intelligent vehicle is too large,the improved distance factor and repulsion coefficient is given by this paper to solve the shortcomings of the unreachable target and the traditional distance factor.Aiming at local optimal conditions,the reasons for the existence of local optimal are analyzed.The repulsive deflection model is introduced first,if the problem is not solved,then a virtual target point is set up until the intelligent vehicle gets rid of the local optimal state.Experiments show that the proposed algorithm can effectively deal with dynamic obstacle avoidance and avoid unreachable targets and local optimal problems.Finally,according to the map information,each intersection is set as a node,and the shortest path is obtained by using the Floyd algorithm according to the travel cost between the intersections and the given starting node;Based on information such as vanishing point and lane line pose,the yaw angle and lateral offset based on the lane centerline are obtained,based on the kinematic model of the intelligent vehicle,the speed control of the intelligent vehicle is realized,and the autonomous navigation experiment of the intelligent vehicle based on machine vision is finally completed.