The machine learning method of phase extraction in interferometry

Abstract Phase extraction of interferometry is a crucial step of optical measurement. In this paper, a machine learning method is proposed to extract the phase from the interferometric signal with the least squares support vector machine, which eliminates the phase unwrapping procedure and enhances the phase measuring accuracy greatly when compared with traditional phase unwrapping methods. Furthermore, it can also work well in the undersampling situation to some extent. Our method is capable of extracting both linear and nonlinear phases. A Michelson interferometer was constructed to demonstrate the validity of the method proposed, which can be potentially applied in many significant fields.