Subspace analysis based machine learning method for automated defect detection from fringe patterns

Abstract

We propose a reliable and robust defect detection method from the noisy fringe patterns obtained in optical interferometry. The proposed method relies on a naive Bayes classifier based machine learning model. The model utilizes the phase derivatives computed using fringe signal subspace analysis as feature vectors. This allows for automated defect identification without the requirement of selection of manual threshold parameters. The simulation analysis of various types of defects corroborates the utility of the proposed method. Further, the method is tested on experimental fringes obtained in diffraction phase microscopy to validate its practical applicability.