[Paper Review] ACSC: Automatic Calibration for Non-repetitive Scanning Solid-State LiDAR and Camera Systems

Jiahe Cui, Jianwei Niu+3 more|arXiv (Cornell University)|Nov 17, 2020

Robotics and Sensor-Based LocalizationEngineering26 references34 citations

TL;DR

ACSC proposes a fully automatic extrinsic calibration method for SSL-camera systems by leveraging time-domain integration and reflectance-based 3D corner estimation of a printed checkerboard, followed by target-based extrinsic estimation.

ABSTRACT

Recently, the rapid development of Solid-State LiDAR (SSL) enables low-cost and efficient obtainment of 3D point clouds from the environment, which has inspired a large quantity of studies and applications. However, the non-uniformity of its scanning pattern, and the inconsistency of the ranging error distribution bring challenges to its calibration task. In this paper, we proposed a fully automatic calibration method for the non-repetitive scanning SSL and camera systems. First, a temporal-spatial-based geometric feature refinement method is presented, to extract effective features from SSL point clouds; then, the 3D corners of the calibration target (a printed checkerboard) are estimated with the reflectance distribution of points. Based on the above, a target-based extrinsic calibration method is finally proposed. We evaluate the proposed method on different types of LiDAR and camera sensor combinations in real conditions, and achieve accuracy and robustness calibration results. The code is available at https://github.com/HViktorTsoi/ACSC.git .

Motivation & Objective

• Address the challenge of extrinsic calibration for SSL-camera systems with non-uniform scan patterns and non-uniform ranging errors.
• Develop a fully automatic, target-based calibration workflow using a printed checkerboard.
• Leverage time-domain integration and reflectance distribution to robustly estimate 3D checkerboard corners from SSL point clouds.
• Integrate 3D corner estimates with 2D image corners to solve the extrinsic parameters.
• Evaluate robustness and accuracy across multiple SSL and camera models in real-world conditions.

Proposed method

• Time-domain integration to densify SSL point clouds while removing frame-level outliers.
• Feature refinement including plane fitting with iterative RANSAC and projection of refined points onto an ideal plane to correct axial ranging noise.
• 3D corner estimation by aligning SSL-derived checkerboard points to a standard model using reflectance distribution as a similarity measure and nonlinear optimization (L-BFGS) to solve for rotation and translation.
• 2D corner detection from camera images using standard checkerboard detection.
• Iterative extrinsic parameter solving: initial PnP via RANSAC, then outlier rejection based on reprojection error, repeating until errors fall below threshold.
• Evaluation of extrinsic calibration accuracy using normalized reprojection error (NRE) and visualization of projection results.

Experimental results

Research questions

• RQ1 Can automatic SSL-camera extrinsic calibration be achieved reliably with non-repetitive SSL scanning without manual intervention?
• RQ2 How can time-domain data fusion improve feature extraction from SSL point clouds for calibration?
• RQ3 Does reflectance-based 3D corner estimation improve 3D-2D corner correspondence for calibration compared to traditional methods?
• RQ4 What is the comparative accuracy of ACSC across different SSL models and camera setups?
• RQ5 What factors (sensor placement, frame integration, target poses) most influence reprojection accuracy in SSL-camera calibration?

Key findings

• ACSC achieves accurate extrinsic calibration across several SSL models (Livox Horizon, Mid-40, Mid-100) and camera pairs.
• Time-domain integration plus feature refinement yields denser, noise-reduced checkerboard measurements from non-repetitive SSL scans.
• Reflectance distribution-based 3D corner estimation enables robust corner extraction and accurate alignment to a standard checkerboard model.
• Quantitative results show superior average reprojection performance compared to prior methods (lower AVG) and high percentages of corners with reprojection error below selected thresholds.
• Increasing the number of target placements and longer integration improves calibration robustness and accuracy.