Simultaneous Intrinsic and Extrinsic Calibration of a Visual-Odometric Sensor System

Abstract

The navigation system of a mobile robot is usually a combination of different sensors. In order to align these measurements, the intrinsic parameters of each sensor and the spatio-temporal extrinsic parameters among them are required to be precisely calibrated. In this paper, a simultaneous intrinsic and extrinsic calibration algorithm of a visual-odometric system is proposed. An initialization process is designed to generate rough estimation of the camera intrinsics, the odometric parameters, and the spatio extrinsic parameters. Then, a joint optimization problem is iteratively solved to obtain a refined calibration result. By taking temporal delay into consideration during the calibration process, the estimation accuracy is highly improved. Experiments are conducted with simulation datasets and also an industrial autonomous guided vehicle (AGV) system. The results are presented to validate the efficiency of our algorithm.