Construction Of Vehicle Information MAP Based On Fusion Of Image And Lidar Information

At present,intelligent driving technology is developing rapidly.Due to the problem of environment awareness and information processing,various environments can not be understood automatically.Constructing vehicle information map based on sensor information fusion method can improve the accuracy of vehicle recognition and it is of great significance to the safe driving of intelligent vehicles.Firstly,the camera and lidar are calibrated jointly.The camera calibration model is established and the internal and external parameters of the camera are optimized by LM gradient descent algorithm.At the same time,the calibration model of lidar is established,and the parameters of lidar are determined by particle swarm optimization,DBSCAN clustering algorithm with variable neighborhood radius and least square method.Since the camera model and the lidar model are calibrated in the vehicle coordinate system,the camera and the lidar can be calibrated jointly through the derivation.Secondly,the camera image and lidar data are preprocessed,and vehicles are identified.The camera color image is processed by the grayscale algorithm of information entropy theory,the filter algorithm based on weight coefficient and the binary algorithm of image segmentation to realize the separation of lane and vehicle shadow.Based on the lane line,the black pixel search method is used to find the intersection line between and the road and the shadow under the vehicle,and then to identify the vehicle.For the lidar sensor,the lidar data is filtered by four filtering methods except the vehicle target.The laser point cloud is clustered by DBSCAN clustering algorithm with variable neighborhood radius to determine the position of the vehicle.Thirdly,the vehicle is identified and the vehicle information map is constructed based on the information fusion of the camera and lidar sensor.Thecubic spline interpolation is used to realize the time registration between the camera information and the lidar information.The information of camera and lidar is fused by coordinate correlation degree feature to ensure that both of them can recognize the same vehicle.The transverse distance of vehicle in image and the longitudinal distance of vehicle detected by lidar sensor are analyzed,and the vehicle information map is constructed by grid method.Finally,the parameters of the cubic Bezier equation are solved by setting the initial conditions of the vehicle.The simulated trajectory of the cubic Bezier curve is plotted by MATLAB software.An experimental vehicle with four cameras and a lidar sensor is used to verify the vehicle information map.The rationality of the information map construction is verified by comparing the simulated track with the experimental vehicle track.