Phase data unwrapping method based on residual error self-encoding neural network

Abstract

The invention discloses a phase data unwrapping method based on a residual error self-encoding neural network. The method comprises the steps: reading generated experiment data, regarding a phase dataunwrapping task as a multi-classification problem, and building the residual error self-encoding neural network; a Zernike polynomial is adopted to generate a simulation initial phase data set, the simulation initial phase data set is wrapped, and a residual error self-encoding neural network is trained; evaluating the trained network model, if a preset precision requirement is met, predicting and classifying the package phases through a residual error self-encoding neural network to obtain a package multiple distribution diagram corresponding to the package phases, and processing a result byusing a two-dimensional median filter to obtain a denoised package multiple distribution diagram; and performing summation operation on the phase wrapping multiple distribution diagram, the wrapped phase data and the to-be-measured data Xtrest to obtain a final unwrapped phase result, and representing fluctuation information of the surface contour of the to-be-measured object. According to the invention, unwrapping operation of various free-form surface phase data can be realized.

Description

technical field [0001] The invention belongs to the technical field of phase data processing in the field of three-dimensional contour precision measurement, and in particular relates to a method for unwrapping phase data based on a residual autoencoder neural network. Background technique [0002] Three-dimensional profile measurement technology is widely used in industrial manufacturing, national defense, aerospace, civil consumption and other fields. Methods such as interferometry and fringe projection belong to the mainstream three-dimensional contour precision measurement technology. During its implementation, it is necessary to collect multiple interferograms through phase shifting to extract phase data, and finally obtain the relevant surface shape measurement results. The phase shift technology will introduce the arctangent operation, and the periodic function atan2 is often used in the computer instead of the arctan function, and its value range is [-π, π]. Theref...

AI Technical Summary

Problems solved by technology

But its computational efficiency is low

Images