基于LM算法的三维点云与二维图像标定方法OA

The insufficient calibration accuracy between LiDAR and camera detection leads to errors in the spatial alignment of LiDAR point clouds and camera images,affecting subsequent feature matching,object detection,and 3D reconstruction accuracy.Therefore,a calibration method based on LiDAR 3D point clouds and monocular camera 2D images is proposed.The method aims to achieve precise detection of large-scale objects and 3D environment reconstruction.In the method,multi-frame point cloud data accumulation is used to obtain relatively dense point cloud measurements and a corner detection algorithm is applied to detect feature corners in the images.Subsequently,the parameters are solved with partial least squares(PLS).Finally,the Levenberg-Marquardt(LM)iterative algorithm is employed to minimize the reprojection error,thereby improving calibration accuracy.The calibration results show that the reprojection error of the SPAAM algorithm is reduced by 8.6%in comparison with that of the classical methods,while the proposed method achieves a nearly 38.2%reduction in reprojection error in comparison with that of the classical methods,validating its accuracy and effectiveness.