An Extrinsic Calibration And Optimization Method Of LiDAR And Camera

The multi-sensor data fusion technology is critical in driverless vehicles.It is a vital connection between the physical world and the unmanned vehicle,and a prerequisite for autonomous driving.The integration of LiDAR knowledge and camera visual information,which is commonly utilized in many robotic applications,has played a key role in unmanned vehicle localization,environmental perception,target tracking,and route planning.The experimental unmanned vehicle platform in this thesis is composed of an LiDAR and a camera.The main work of this thesis is to complete the calibration of the extrinsic parameters of the LiDAR and the camera in the unmanned vehicle experimental platform and the intrinsic parameters calibration of the camera.The main research work of this thesis is as follows.(1)The research on LiDAR and camera calibration method and the key technical points has been introduced.The experimental platform in the thesis has been designed to complete the multiple sensor calibration system.(2)Research on the intrinsic parameters calibration method of the camera.Compared with several traditional methods of feature point detection,combined with the advantages of traditional algorithms,an efficient and stable feature point detection algorithm has been designed with consistent feature patterns of the calibration board.Once the image feature points has been detected,a camera coordinate system for the captured calibration images can be established.The homography matrix can be calculated using the homography of the space corresponding to the plane,as well as the homography matrix is decomposed to obtain the pose of the camera calibration board in the camera coordinate system and the camera intrinsic parameters.Finally,the smallest reprojection error has been utilized to optimize the camera intrinsic parameters.(3)Research on calibration methods of LiDAR and camera extrinsic parameters.First,the corresponding frames of the LiDAR and the camera data need to be extracted.Then,the3 D points attached to the calibration board from the optical radar point cloud data need to be filtered out,and we can estimate the pose of the calibration board in the LiDAR coordinate system.Simultaneously,we use the corner detection algorithm to calculate the localization of corners in the images,and the pose of the calibration board in the camera coordinate system can be estimated with the localization of the corner points in the pixel coordinate.Finally,we use the pose relationship of the calibration board between the LiDAR and the camera coordinate system to construct the coordinate system transformation equation and solve the rotation matrix and translation vector between the sensor coordinate systems.