Self-calibration of a vehicle's acquisition system with cameras, IMU and 3D lidar

摘要

This article presents a fully automated calibration method suitable for complex acquisition systems made with one or more cameras, an inertial measurement unit and a 3D lidar. The principle consists in estimating the intrinsic and extrinsic parameters by matching features detected in the camera images with the 3D point cloud provided by the rangefinder. This work proposes a mathematical formalization for the unification of the three types of sensors within the same likelihood function and a minimization algorithm with three families of parameters that enables the simultaneous estimation of all calibration parameters. Experiments conducted on synthetic acquisition systems and real sequences are presented and assess the area of convergence of the proposed approach and its performance in terms of accuracy, correctness and robustness in the presence of noise.