Galibr: Targetless LiDAR-Camera Extrinsic Calibration Method via Ground Plane Initialization

TL;DR

Galibr tackles automatic, targetless LiDAR-camera extrinsic calibration in unstructured outdoor environments by introducing a two-stage pipeline that first uses ground-plane information (GP-init) to generate robust initial sensor poses and then refines the extrinsics via edge-matching optimization. The method leverages ground-plane features with Structure-from-Motion and RANSAC for the camera, and a constant-velocity IESKF with TRAVEL segmentation for the LiDAR, followed by an edge-projection-based Levenberg–Marquardt optimization using ELSED edges. Empirical results on KITTI and KAIST datasets show improved accuracy, especially when GP-init is used, and faster runtimes compared with state-of-the-art methods, highlighting practical applicability for moving ground vehicles. The targetless, ground-plane–driven approach offers robust calibration without dedicated targets, enabling maintenance-friendly deployment across diverse terrains and sensor configurations; future work includes integrating IMU data and adaptive filtering to further enhance performance.

Abstract

With the rapid development of autonomous driving and SLAM technology, the performance of autonomous systems using multimodal sensors highly relies on accurate extrinsic calibration. Addressing the need for a convenient, maintenance-friendly calibration process in any natural environment, this paper introduces Galibr, a fully automatic targetless LiDAR-camera extrinsic calibration tool designed for ground vehicle platforms in any natural setting. The method utilizes the ground planes and edge information from both LiDAR and camera inputs, streamlining the calibration process. It encompasses two main steps: an initial pose estimation algorithm based on ground planes (GP-init), and a refinement phase through edge extraction and matching. Our approach significantly enhances calibration performance, primarily attributed to our novel initial pose estimation method, as demonstrated in unstructured natural environments, including on the KITTI dataset and the KAIST quadruped dataset.

Figures (8)

• Figure 1: Overview of Galibr. Galibr estimates the LiDAR-camera extrinsic calibration result in two steps: initial relative pose estimation using a ground plane and edge matching-based extrinsic calibration.
• Figure 2: System overview of our approach. Unlike existing LiDAR and camera extrinsic calibration methods, which generally need initial values, our approach focuses on estimating initial relative pose using ground plane features. With the two-step estimation, initial pose estimation step and extrinsic calibration step, our approach outputs more accurate and robust extrinsic calibration results.
• Figure 3: (a) Ground feature extraction result (red points) using SfM. (b) Ground feature extraction result (black points) using LiDAR odometry and TRAVEL oh2022ral. Using ground extraction results, we estimate the ground plane and the relative pose from the ground for the initial relative pose estimation of two sensors.
• Figure 4: The edge detection examples in images and object extraction examples in LiDAR point clouds.
• Figure 5: Visual description of different edge points of a foreground object from different views of LiDAR and camera.