Targetless Automatic Edge-to-Edge Extrinsic Calibration of LiDAR-Camera System Based on Pure Laser Reflectivity

Abstract

High-precision calibration of LiDAR and cameras is an essential precondition for multisensor fusion systems. Since controlled targets are not always available in field, targetless calibration methods have emerged as a practical alternative to traditional target-based approaches. However, the complexity and variability of natural scenes pose significant challenges to feature extraction and matching in the targetless calibration process. To address these issues, this article proposes a targetless, automatic edge-to-edge calibration method based solely on laser reflectivity, aimed at enhancing calibration accuracy and feature consistency in the LiDAR-camera calibration (LCC) task. Traditional methods primarily focus on the 3-D information of LiDAR point clouds for feature detection. In contrast, the proposed method leverages the superior capabilities of laser reflectivity to perceive color, depth, and material properties. This strategy enables robust edge feature consistency in 2-D, thereby enhancing overall calibration accuracy. Experimental results demonstrate that the proposed method reduces translation error by an average 61.9% compared to existing state-of-the-art targetless approaches, offering enhanced feature extraction accuracy and robustness, particularly in complex and unstructured environments.