A Method for Synchronous Automated Extrinsic Calibration of LiDAR and Cameras Based on a Circular Calibration Board

Abstract

The fusion of LiDAR and camera data is a promising approach for improving the environmental perception and recognition abilities of robots. The fusion of data from the two sensors plays a vital role in enhancing robotic localization capabilities. Current methods of data fusion first involve estimating the intrinsic parameters of cameras and then developing a transformation matrix between recordings and LiDAR frames. However, the drawback of these methods is the accumulation of errors. To improve the data fusion accuracy, we propose a novel approach for performing the automatic extrinsic calibration between a LiDAR system and a camera utilizing a specially designed circular calibration board. The method consists of the following steps. First, the central feature and the plane normal to the calibration board in the point clouds and images are extracted. Second, constraint equations based on the corresponding features are established to obtain an initial estimate of the transformation matrix between the LiDAR and camera. Finally, the triple factorization-based symmetric and nonnegative latent factor model is expanded to obtain reliable results in extrinsic calibration. Simulations and experiments indicate that the proposed method aligns the 3-D points in the LiDAR frame with the pixels in the image frame with 47.59% higher accuracy than the state of the art. Even under the influence of noise at a level of 0.1 dB, the maximum reprojection error remains below 9.35 pixels, thereby further underscoring the robustness of the proposed method in the face of noise challenges.