DescriptionHolistic calibration algorithms of stereo deflectometry are generally based on iterative computation and therefore heavily sensitive to the accuracy of input initial values. Image distortion seriously affects the estimation accuracy of the initial value, which causes the iterative computation to converge in the wrong position or unable to converge. An algorithm is presented to increase the accuracy and robustness of the calibration of a stereo deflectometry system. Firstly, based on the knowledge that distortion of the pixels which close to image center can be ignored, a search window with a certain border length is placed at image center. The border size of the window can be determined according to the proposed algorithm and experimental results. Secondly, considering distortion center and image center cannot normally coincide in an actual imaging system, the search window is moved within the image sequentially. Accurate initial values are searched with the window moving, and determined by making the proposed evaluation function reach the minimum. Experimental results affirm the proposed algorithm can significantly improve the calibration accuracy of a stereo deflectometry system. The RMS (root mean square) of calibration error can be reduced to 0.05 pixels from 0.31 pixels by using the proposed algorithm.
|7th International Conference on Optical and Photonic Engineering
|Degree of Recognition