Planar Visual Metrology using Partition-based Camera Calibration

Abstract

In visual metrology the real imaging procedure is a complicate nonlinear system. Even after rectification of the radial distortion of the lens, it is hard to achieve high accuracy with single group of planar extrinsic parameters. According to the characteristics of the radial distortion of lens, a partition-based camera extrinsic calibration method for planar visual measurement is proposed. The image plane is partitioned into 9 subspaces according to the lens radial distortion, and in each subspace one group of extrinsic parameters of objects plane is calibrated separately. Although there is only one objects plane where the visual measurement is to be done, the groups of extrinsic parameters calibrated in each subspace are different because of the nonlinear in real system. In the subspace that the image points belong to, the corresponding group of extrinsic parameters of the subspace is adopted to reconstruct the 3D positions of the objects in the camera coordinate system. Finally experimental results show that the partition-based calibration method performs well in visual metrology. Compared with the method which uses single group of extrinsic parameters the proposed method can achieve higher accuracy in measuring distance of planar objects from their perspective images.