High-Accuracy Calibration of On-Site Multi-Vision Sensors Based on Flexible and Optimal 3D Field

Global calibration of multi-vision sensors in the railway fields is easily affected by on-site complex environments, such as lighting and self-occlusion, which makes it difficult for existing methods to achieve high-accuracy calibration. In this paper, a high-accuracy and flexible calibration method of multi-vision sensors in the outdoor railway fields based on flexible and optimal 3D data field through the combination of articular arm and metal target embedded with luminescent LEDs is proposed. Firstly, a high-accuracy and flexible 3D data field with multiple angles and views is constructed by the articular arm and the metal target rapidly; Secondly, the high-frequency multi-exposure imaging mode and multi-scale image feature points extraction are adopted, and the high-accuracy reposition is achieved through Kalman filter, which can reduce the impact of image noise efficiently; Finally, the RANSAC method is utilized to optimize the 3D data field, forming the optimal 3D data field, and the optimal target chains corresponding to each group of cameras are established. The maximum likelihood solution of global parameters is solved by nonlinear optimization. Simulation experiments verify the feasibility of the proposed method. Meanwhile, physical experiment results show this method can reduce the outdoor environment impact and improve the calibration and measurement precision effectively. In addition, the computational efficiency of the proposed method is about 11.2 times than multi-images averaging method, the calibration and measurement accuracy improve about 20 and 2.9 times respectively, which is suitable for the high-accuracy and flexible global calibration in the complex railway environment.