Improvement to LiDAR-camera extrinsic calibration by using 3D-3D correspondences

초록

In data fusion, calibration between 3D light detection and ranging (LiDAR) systems and cameras is one of the most fundamental tasks. We propose a calibration method to improve calibration accuracy by using a planar board. Most previous studies have attempted to find correspondences between 3D point clouds and 2D image pixels. In this study, a rigid transformation between LiDAR and camera was performed by exploiting the 3D-3D vertices and plane correspondences. First, the features, vertices and plane, are extracted from the camera and LiDAR, respectively. By compensating for the missing part of the 3D point cloud at the edges of the board, we improve the accuracy of the vertex estimation instead of only using the traditional RANSAC algorithm. Second, the rigid transformation is estimated by matching these 3D correspondences and refined by an ICP improvement. Finally, to evaluate the robustness of the proposed calibration method, we performed experiments and analyzed the calibration errors in both simulated data and real datasets.

키워드