Single-Shot ROI-Based Targetless LiDAR-Camera Calibration via Edge Features

The integration of 3D LiDAR scanners and 2D cameras through sensor fusion is widely adopted in robotics, as it leverages the complementary strengths of both modalities. Accurate fusion requires precise estimation of the six degrees-of-freedom (6-DOF) extrinsic transformation between sensors. Manual measurement is prone to significant errors, while extrinsic parameters may drift over time due to environmental factors or mechanical shifts, leading to degraded system performance. Prior research has demonstrated targetless calibration approaches, often relying on large datasets or extended sequences. In contrast, this study investigates a lightweight strategy that performs calibration using a single frame and a region of interest (ROI) from arbitrary environments. The proposed approach exploits edge features extracted from both LiDAR point clouds and camera images, with optimization carried out via grid search. Experimental evaluations on the KITTI dataset demonstrate that the method can effectively recover extrinsic parameters under various perturbations of the calibration matrix.