An improved geometrical calibration method for stereo deflectometry by using speckle pattern

Abstract

System geometrical calibration is critical for accurate stereo deflectometry. Conventional calibration approaches involve displaying checkerboard, circle, or fringe patterns on a systemic LCD screen, and the system parameters are calculated via reflection using a markerless mirror at three different locations. These approaches introduce form errors into the system because of defocus images. When using fringe patterns, the calibration procedure becomes more complex for acquiring a plethora of fringe patterns in two orthogonal directions, and ambient light influences the accuracy of the unwrapped phase map. In this study, we propose a novel calibration method for a stereo deflectometry system using a speckle pattern to improve the system measurement accuracy with a flexible calibration procedure. The proposed method includes an improved similarity evaluation function based on KAZE features to improve the accuracy of feature matching between images with perspective transformation, and a coarse-to-precise algorithm is applied to suppress the defocusing effect and refine the matched set. The initial geometrical parameters are calculated based on the refined matching results. The final procedure optimises all geometrical parameters into a single cost function using the Levenberg–Marquardt algorithm. Experiments using a stereo deflectometry system demonstrated that the proposed method achieves high accuracy with while being simple and flexible.