Single-frame fringe analysis method based on multi-scale generative adversarial neural network

Abstract

The invention discloses a single-frame stripe analysis method based on a multi-scale generative adversarial neural network. The method comprises the following steps: constructing a multi-scale generative adversarial neural network model; constructing a comprehensive loss function L of the multi-scale generative adversarial neural network model; collecting training data of the multi-scale generative adversarial neural network, and training the multi-scale generative adversarial neural network by using the training data; and inputting a to-be-measured stripe image into the trained multi-scale image generator to obtain a corresponding sine term, cosine term and modulation degree graph, and calculating a phase by using an arc tangent function. After the neural network is trained, complex calculation parameters do not need to be manually set in the calculation process, and operation is easier and more convenient. Due to the fact that the input of the neural network is a single stripe image,an efficient and high-precision phase calculation method is provided for stripe analysis of a moving object.

Description

technical field [0001] The invention belongs to the technical field of optical measurement, and specifically relates to a single-frame fringe analysis method based on a multi-scale generation confrontational neural network. Background technique [0002] With the advancement of computer technology, information technology and optoelectronic technology, optical three-dimensional measurement technology has developed rapidly. Optical three-dimensional measurement technology is a technology based on modern optics, integrating optoelectronics, signal processing, image processing, computer graphics, pattern recognition and other science and technology. It uses optical images as a means and carrier for detecting and transmitting information. Its purpose is to extract useful signals from images and complete the reconstruction of three-dimensional solid models. Optical 3D measurement technology is usually divided into two categories according to different imaging and illumination meth...

AI Technical Summary

Problems solved by technology

However, the disadvantage is that due to the need to collect a series of fringe images for analysis, the measurement efficiency is low, and it is difficult to meet the contour measurement of moving targets.

However, the disadvantage of this method is that the implementation process is relatively complicated, and the process of parameter adjustment is relatively cumbersome, and the time cost of phase analysis is very high

Images