A novel camera calibration method based on known rotations and translations
Zhangfei Chen,Xuelong Si,Dan Wu,Fengnian Tian,Zhenxing Zheng,Renfu Li
DOI: https://doi.org/10.1016/j.cviu.2024.103996
IF: 4.886
2024-03-29
Computer Vision and Image Understanding
Abstract:It is often difficult to align the camera optical center with the rotation center of the motion mechanism during active vision calibration, and this dis-alignment could lead to inaccuracy of the constraint equations and calibration results. To circumvent such issues, this paper proposes a novel method for active vision camera calibration and facilitates its real-world implementation. In this method, the rotational motion axis is assumed not necessarily to pass through the camera's optical center, and the constraint equations are established by correlating image matched points before and after camera motions. The displacements caused by rotation are incorporated into the constraint equations to improve the accuracy of calibration. Numerical simulation and experiments are conducted. The feasibility of this proposed method is justified by using both synthetic and experimental data, and the results show that this method could estimate camera intrinsic parameters with high accuracy. Furthermore, it is found that small rotation angles can better the calibration accuracy even under the influence of external noises. This study suggests that this novel method has great potential for visual and image applications.
computer science, artificial intelligence,engineering, electrical & electronic