Camera-based optical metrology crucially relies on a proper identification of intrinsic (optical center position) and extrinsic (camera position) parameters of the used camera. A novel approach for processing phase data from multiple views of a target grid is presented, allowing these identifications within a pinhole camera model. First, the homography associated with the perspective distortion of each grid image is accurately identified using a phase-locking loop. Then, the affine transform for each view is determined, with the constraint of an identical set of intrinsic camera parameters. This requires slightly adjusting each homography. As there are highly redundant data, a criterion has to be chosen to optimize the result. The chosen criterion is the simultaneous minimization of the standard deviation between in-plane grid line displacements between the acquired grid images and the reconstructed ones. Experimental results demonstrate the efficiency of the method. (C) 2017 Optical Society of America
展开▼
