Spatiotemporal LiDAR-IMU-Camera Calibration: A Targetless and IMU-Centric Approach Based on Continuous-time Batch Optimization

Sensor calibration is one of the basic tasks for a multimodal sensing system. In this paper, we present a two-stage spatiotemporal calibration method for a common-used sensor suite, i.e. the LiDAR-IMU-camera sensor combination. In the first stage, the proposed method combines correlation analysis with hand-eye calibration to acquire the initial values of the time offset and the extrinsic rotation parameters for IMU-LiDAR and IMU-camera respectively, and refines the time offset through aligning the rotational trajectories. In the second stage, continuous-time batch optimization framework is employed to jointly estimate the extrinsic parameters of IMU-LiDAR and the trajectory of the IMU, and the optimized trajectory is utilized to calibrate the IMU-camera extrinsic parameters. In addition, the data association scheme is designed according to the modality of the LiDAR and the camera rather than the assistance of artificial targets. We validate the proposed method on both public dataset and self-record data. Experimental results show the feasibility and the performance of the proposed method through comparision with the state-of-the-art calibration methods for IMU-LiDAR and IMU-camera calibration.