Objectives Accurate and stable extrinsic parameter calibrations are the basis of fusing perception and positioning of cameras and light detection and ranging (LiDAR). To address the problems of the Dhall calibration method in constraint accuracy, stability, and convergence, this paper proposed a LiDAR and camera calibration method with joint point-line-plane constraints.
Methods First, line-line and plane-plane constraints were added, and the Kabsch method was used to obtain a closed-form solution, which reduced the influence of noise in the single corner constraint and improved the accuracy and stability of the calibration results. Then, multi-frame point clouds were superimposed to fit lines, and the fitting corners were averaged, which further reduced the influence of the point cloud noise, thus improving the convergence speed and stability of the calibration method.
Results Experimental results show that the proposed method can converge within 20 frames, and the reprojection error of the corner points was less than 2 cm.
Conclusions The proposed method has faster convergence speed and higher accuracy than the original method.