Automatic Multi-camera Multi-LiDAR Extrinsic Calibration using Geometric Corners

Abstract

In autonomous driving systems, multiple sensors are installed to provide complementary information. Extrinsic calibration is the prerequisite of multi-sensor fusion, which is critical to the accuracy and performance of the fused system. In this paper, we propose a unified algorithm for extrinsic calibration among multiple cameras and light detection and ranging (LiDAR) sensors using only a standard checkerboard. The geometric information of the checkerboard, containing plane coefficients, boundary points and geometric corners, is extracted automatically with a lightweight approach by leveraging the characteristics of mechanical LiDARs. A coarse-to-fine manner is adopted in the extrinsic calibration by utilizing these features in different stages. Our algorithm has been experimentally validated through simulations and real-world experiments. The experimental results show that the proposed method is capable of providing accurate extrinsic parameters and is robust in different levels of Gaussian noise. All procedures in our approach operate automatically, without any manual intervention, which is beneficial to practical applications.