LIP-Calib: Automatic Targetless Extrinsic Calibration of LiDAR-Camera System Based on Low Intensity Perception

Robust and accurate LiDAR–Camera Calibration (LCC) is a fundamental prerequisite for environmental perception in multi-sensor fusion systems. Targetless LCC has recently emerged as a practical alternative to target-based approaches, as it eliminates the need for fabricating controlled targets while achieving comparable accuracy. However, the reliance on dense point clouds not only increases computational overhead but also amplifies noise due to laser beam divergence, further degrading edge quality and hindering the generalization of targetless methods to natural scenes. In addition, variations in camera intrinsics significantly affect the density distribution of LiDAR projection maps, while most existing methods are tailored to fixed sensor setups, limiting their adaptability to diverse hardware configurations in real-world deployments. To address these issues, LIP-Calib (Low Intensity Perception Calibration) is proposed. A LiDAR intensity perception strategy is devised to selectively retain low-intensity points that effectively capture structural contours in natural scenes, thereby enhancing the quality of edge extraction. In parallel, an energy-aware binarization scheme mitigates the impact of laser projection density variations induced by camera intrinsics, ensuring strong adaptability across diverse hardware configurations. Experimental results demonstrate that the proposed method achieves superior calibration accuracy compared to state-of-the-art targetless approaches, with performance comparable to or exceeding that of target-based methods. In addition to high accuracy, the proposed method demonstrates strong robustness to variations in camera intrinsics and point cloud distributions, as validated through extensive tests across five types of cameras and two types of LiDAR sensors.