Stable Minimum Solution Calibration Method Using 2-D Laser Radar and Camera

• Published in: IEEE sensors journal Vol. 24; no. 3; pp. 3711 - 3721
• Main Authors: Peng, Meng, Wan, Qin, Zeng, Saifeng, He, Jiale
• Format: Journal Article
• Language: English
• Published: New York The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 01.02.2024
• Subjects: Accuracy; Algorithms; Calibration; Cameras; Constraint modelling; Error analysis; Lasers; Lidar; Noise levels; Numerical stability; Obstacle avoidance; Robustness (mathematics); Simultaneous localization and mapping; Stability; Target recognition; Unmanned vehicles
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2023.3342913

The 2-D laser radar and cameras are widely used to perceive external environments in unmanned vehicle navigation, target recognition, mobile robot obstacle avoidance, and simultaneous localization and mapping. Herein, the two sensors must be calibrated first to fuse laser data and image. The minimum solution method for calibration is one of the most popular ones but has some shortcomings, such as poor numerical stability and accuracy. Therefore, we present a new stable minimum solution calibration method, which introduces the robust perspective three-point algorithm (RP3P) to solve the perspective three-point problem (P3P) and improve the mechanisms of optimal solution selection. First, according to the P3P constructed using three checkerboards, the enhanced RP3P algorithm is used to improve the stability of the solution. Second, an error measurement model based on region and boundary constraints is proposed to evaluate the deviation of the multiple solutions to select the optimal solution. Simulation and actual experimental results show that the proposed algorithm can significantly improve the percentage of valid solutions and calibration accuracy compared with the algorithms in the literature. In the simulation experiments, compared with Francisco’s, Hu’s, and Zhou’s methods, the percentage of valid solutions of our method is improved by 7%–33%, 6%–10%, and 5%–9% under different noise levels. Furthermore, the accuracy of the calibration result of our calibration method obtained significant performance improvements.