Online Extrinsic Parameter Calibration for Robotic Camera–Encoder System

Abstract

Cameras and encoders are widely used in mobile robots, and extrinsic parameter calibration of these sensors is crucial in practical performance. The existing approaches mainly rely on manual measurements, accurate computer-aided design (CAD) models, or carefully designed artificial landmarks. This paper presents a novel approach for automatically calibrating the extrinsic parameters of the robotic camera-encoder system. The approach first calculates a coarse estimation of the external parameters as well as the scale of the visual system, via free-scale hand-eye calibration of the camera and odometer. However, the coarse calibration result and scale of the visual system do not satisfy the accuracy requirement for further mobile robots' applications such as localization and navigation. A nonlinear optimization algorithm that considers both bundle adjustment and odometer measurement error functions is developed to refine the extrinsic parameter calibration result. This coarse-fine approach is computationally efficient and can achieve online calibration during the vehicle motion automatically. Furthermore, it can realize the calibration without using any artificial landmark or prior knowledge about CAD models. Finally, comparisons to other classic calibration approaches are performed with a series of simulations and experiments to illustrate the effectiveness of the approach.