RAVES-Calib: Robust, Accurate and Versatile Extrinsic Self Calibration Using Optimal Geometric Features

TL;DR

RAVES-Calib tackles targetless LiDAR–camera extrinsic calibration across diverse sensors by leveraging Gluestick to obtain robust 2D–3D correspondences for a rough initial estimate and by analyzing feature distribution to adapt feature weighting. It introduces a multi-type feature set (points, lines, and depth-continuous edges) and a Jacobian/Hessian-based contribution index to guide a weighted least-squares optimization without requiring an initial transform. Extensive indoor/outdoor experiments with solid-state and spinning LiDARs, and full-frame and pinhole cameras demonstrate state-of-the-art robustness and accuracy, often rivaling target-based methods. The toolkit is open-source and aims to facilitate practical, automated calibration across varied sensing platforms.

Abstract

In this paper, we present a user-friendly LiDAR-camera calibration toolkit that is compatible with various LiDAR and camera sensors and requires only a single pair of laser points and a camera image in targetless environments. Our approach eliminates the need for an initial transform and remains robust even with large positional and rotational LiDAR-camera extrinsic parameters. We employ the Gluestick pipeline to establish 2D-3D point and line feature correspondences for a robust and automatic initial guess. To enhance accuracy, we quantitatively analyze the impact of feature distribution on calibration results and adaptively weight the cost of each feature based on these metrics. As a result, extrinsic parameters are optimized by filtering out the adverse effects of inferior features. We validated our method through extensive experiments across various LiDAR-camera sensors in both indoor and outdoor settings. The results demonstrate that our method provides superior robustness and accuracy compared to SOTA techniques. Our code is open-sourced on GitHub to benefit the community.

Figures (11)

• Figure 1: Colorized LiDAR points by camera RGB image with calibrated extrinsic parameters using the proposed method. The second row is the overall color point cloud, and the first row is the enlarged picture of the local details in the color point cloud.
• Figure 2: The pipeline of LiDAR-camera calibration process.
• Figure 3: Gluestick can find the correspondence between point and line features in LiDAR intensity image and camera images.
• Figure 4: Feature contribution heatmap. The horizontal and vertical axes represent the x and y axes of the image, respectively, in pixels. Green represents high contribution, yellow represents low contribution.
• Figure 5: Multi-type feature projection image. A, Depth-continuous edge features. B, Intensity image point features. C, Intensity image line feature.