A Novel Global Camera Calibration Method Based on Flexible Multidirectional Target

Multiviewpoint or multisensor 3-D vision measurement is widely applied for large-scale measurement. The global coordinates of multiple local measurement coordinates during the measurement process are often unified by using a global camera. However, the parameters of the global camera are easily affected by external factors such as vibrations and lighting, which can reduce the 3-D reconstruction accuracy during the unification process. Therefore, it is necessary to use calibration targets to calibrate the parameters of the global camera. However, due to the field of view (FOV) limitation of the global camera, the size of the calibration target needs to be strictly controlled. Manufacturing large and high-precision targets is difficult. The accuracy of large targets is not easy to guarantee, and using large targets can easily exceed the FOV of the global camera. In contrast, small and high-precision targets only occupy a small part of the image, which leads to a decrease in the accuracy of camera calibration. To solve this problem, a novel camera calibration method based on a flexible multidirectional target (FMT) is proposed. The designed FMT consists of co- planar small targets (CSTs) and a polyhedral skeleton. The FMT is placed in the FOV of the global camera, and target images are collected from multiple angles. An automatic corner extraction algorithm and an optimization algorithm are proposed based on the topological correspondence between the 2-D coordinates of the corners in the image and the 3-D coordinates of the corners in Euclidean space. These algorithms are used to calculate the parameters and distortion coefficients of the global camera. The algorithm is evaluated through real experiments. The experimental results show that using FMT can improve the measurement accuracy by 83% compared with using chessboard targets, confirming the effectiveness of the algorithm.