Automatic and Targetless LiDAR–Camera Extrinsic Calibration Using Edge Alignment
Jun Yin, Fei Yan, Yisha Liu, Yan Zhuang
Abstract
In this article, an automatic and targetless extrinsic calibration method is proposed for the light detection and ranging (LiDAR)–camera system, which can calibrate extrinsic parameters from coarse to fine in natural scenes. The calibration method contains a motion-based stage and a feature-based stage. In the motion-based stage, LiDAR and camera motions are inferred during the localization. Also, we proposed a motion selection algorithm to identify sensor motions with stable scales to simplify the cost function of the calibration process. Then, the initial extrinsic parameters are estimated by a joint optimization. In the subsequent feature-based stage, a novel recalibration algorithm using edge alignment is proposed to refine the translation vector of the extrinsic parameters. A novel framework is designed in this stage to extract LiDAR edge points, and a new cost function is proposed to refine the extrinsic parameters. Experiments on public datasets and real-world scenes show the practicability and validity of our calibration method.