基于深度学习的剪切散斑干涉条纹图滤波方法 林薇.pdf

: .
激光与光电子 30ms 左右便能实
现条纹图的高质量滤波，符合动态无损检测的发展需求，为相位条纹图的噪声滤除提供了新
的思路。
关键词 剪切散斑干涉；相位图；深度学习；噪声；图像处理；无损检测
中图分类号 O439 文献标志码 A

Phase Fringe Patterns Filtering Method for Shearography Based on Deep
Learning
Lin Wei1, Cui Haihua1,*, Zheng Wei2, Zhou Xinfang2, Xu Zhenlong1, Tian Wei1
1 College of Mechanical & Electrical Engineering, Nanjing University of Aeronautics
and Astronautics, Nanjing, 211106;
2 AVIC Xi'an Aircraft Industry Group Company, Ltd, Xi'an, 710089
Abstract As a non-contact high-precision optical full-field measurement method,
shearography can be used for non-destructive detection of internal defects of composite
materials. However, the phase fringe pattern obtained contains a lot of speckle noise, which
will seriously affect the detection results and accuracy. Therefore, a phase fringe filtering
method based on unsupervised image style conversion model CycleGAN was proposed, and
the original noise phase fringe obtained by shearography was converted into an ideal noiseless
fringe by network training, so as to achieve noise filtering in the phase fringe pattern.
Experimental results show that this method can achieve high-efficiency filtering of noise in
areas where the stripe distribution is relatively sparse, with clear boundaries and significant
contrast in filtered images. The running time of the algorithm is obviously better than other
methods, only abo