Geometry-Preserving in 3D Gaussian Splatting for LiDAR-Camera Extrinsic Calibration
For autonomous driving systems requiring accurate LiDAR-camera calibration, this method improves targetless calibration accuracy without manual setup.
Existing 3D Gaussian Splatting-based calibration methods prioritize rendering quality, causing geometric drift from true LiDAR structure. The proposed framework preserves metric geometry via multi-view depth supervision and blocked photometric gradients, outperforming targetless methods on public driving datasets.
Accurate LiDAR-camera calibration is essential for robust multi-modal perception. Targetless approaches avoid manual setup but remain limited by the scarcity of discriminative cross-modal features. Recent methods address this by reconstructing the scene within a differentiable model, enabling extrinsic optimization through dense photometric supervision. Among these, 3D Gaussian Splatting (3DGS) has been widely adopted as a geometric proxy that bridges LiDAR and camera within a single differentiable framework. However, since 3DGS was originally designed for novel view synthesis, existing methods tend to prioritize rendering quality, causing the proxy geometry to drift from the true LiDAR structure. We propose a framework that preserves the metric geometry of the Gaussian proxy by aggregating multi-view LiDAR observations for dense depth supervision and blocking photometric gradients from updating the Gaussian spatial parameters. We validate our method on public driving datasets, where it consistently outperforms existing targetless methods in calibration accuracy.