Joint calibration method for non-scanning LiDAR and camera

Abstract To address the issues of Flash LiDAR performance being susceptible to distance and lighting conditions, as well as the uneven point cloud imaging, caused by its non-scanning characteristics (single-shot planar imaging leading to point cloud interference, ‘uneven distribution, and ‘thickness’ phenomenon’), this paper proposes a joint calibration method for Flash LiDAR and camera. Currently, there is a relative scarcity of research on the joint calibration of non-scanning LiDAR with camera.This method is based on target-based calibration framework, and in response to the aforementioned non-scanning characteristics, constructs a complete technical chain of ‘point cloud thickness elimination - corner point guided matching - adaptive optimization’: through point cloud preprocessing to eliminate ‘thickness’ to extract effective target point clouds, an improved corner point estimation method is used to accurately extract corner points of irregularly distributed checkerboard point clouds, and an adaptive optimization function is designed to calculate the relative transformation relationship between point clouds and image corner points and minimize the reprojection error. Experimental evaluations conducted in real measurement environments demonstrate that the average reprojection error of this method is approximately 0.39 pixels, providing a theoretical foundation and technical support for subsequent improvements in the accuracy of data fusion between non-scanning LiDAR point clouds and camera images.