Fringe projection time phase unwrapping method based on deep learning

Abstract

The invention discloses a fringe projection time phase unwrapping method based on deep learning. Firstly, four sets of three-step phase shift grating patterns are projected to a to-be-tested object, the frequencies are 1, 8, 32 and 64 respectively, and a camera collects a raster image and obtains a wrapped phase image by using a three-step phase shift method; then, a multi-frequency algorithm based on time phase unwrapping is used for carrying out phase unwrapping on the wrapped phase image to obtain a periodic level map of a phase with the frequency of 64; a residual convolutional neural network is constructed; the input data is set to be the wrapped phase image with the frequencies of 1 and 64, and the output data is the periodic level map of the phase with the frequency of 64; finally,a training set and a verification set are made to train and verify the network; and the network verifies a test set to output the periodic level map of the phase with the frequency of 64. According tothe fringe projection time phase unwrapping method based on deep learning in the invention, the deep learning method is adopted and the wrapped phase image with the frequency of 1 is used for unwrapping the wrapped phase image with the frequency of 64; and an absolute phase image with less error points and higher accuracy can be obtained.

Description

technical field [0001] The invention belongs to the technical field of optical measurement, and in particular relates to a deep learning-based fringe projection time phase expansion method. Background technique [0002] In recent decades, rapid 3D shape measurement technology has been widely used in various fields, such as intelligent monitoring, industrial quality control and 3D face recognition. Among the many three-dimensional shape measurement methods, the fringe projection profilometry based on the principle of structured light and triangulation is one of the most practical technologies. The three-dimensional measurement generally needs to go through three processes, which are phase recovery, phase unwrapping, and phase-to-height mapping. [0003] Among phase recovery techniques, the two most commonly used methods are Fourier profilometry and phase-shift profilometry. Fourier profilometry only needs a fringe pattern to extract the phase, but this method is affected by...

AI Technical Summary

Problems solved by technology

Therefore, for the 3D imaging technology based on fringe projection profilometry, there is still a lack of a method with both measurement accuracy and measurement efficiency.

Images