Deflectometry has been widely used to detect defects on specular surfaces. However, it is still very challenging to detect defects on semispecular or diffuse surfaces because of the low contrast and low signal-to-noise ratio. To address this challenge, we proposed a phase-modulation combined method for accurate defect detection. Based on the phase and modulation of captured fringes, a dual-branch convolutional neural network is employed to simultaneously extract geometric and photometric features from the phase-shifting pattern sequence and modulation, which improves the defect detection performance significantly. Compared to state-of-the-art methods, we believe the results demonstrated the proposed method's effectiveness and capability to reduce false positives. (C) 2020 Optical Society of America
展开▼
