基于生成对抗网络数据扩充的缺陷识别方法  

GANs-based synthetic data augmentation for defects recognition

作　　者：邱根[1] 王锂 白利兵 Qiu Gen;Wang Li;Bai Libing(School of Automation Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China)

机构地区：[1]电子科技大学自动化工程学院,成都611731

出　　处：《电子测量与仪器学报》2021年第2期212-220,共9页Journal of Electronic Measurement and Instrumentation

基　　金：四川省科技计划项目(2020JDR0049);四川省重大科学仪器设备专项(2019ZDZX0045)资助。

摘　　要：可视化无损检测(NDT)在深度学习技术发展下,在数据处理方面正面领着巨大的机遇。但是,获取足够的标记数据集是一个很大的挑战。实现无损检测图像数据集的扩充有利于提升深度学习在缺陷检测中的能力。因此,通过研究无损检测图像数据特点,结合循环一致生成对抗网络(CycleGANs)方法,对现有的数据进行了有效的扩充。改善了深度卷积神经元网络(DCNN)从而有效的利用扩充数据来提升对缺陷图像的识别能力。最后,通过对比实验,展示了本扩充数据对提升缺陷检测网络训练具有重要作用。Visualized non-destructive testing(NDT), with the development of deep learning technology, is leading a huge opportunity in data processing. However, obtaining sufficient labeled data sets is a big challenge. The realization of the expansion of the non-destructive inspection image data set is conducive to improving the ability of deep learning in defect detection. Therefore, this article has effectively expanded the existing data by studying the characteristics of non-destructive testing image data, combined with CycleGANs(CycleGANs) method. Improved the deep convolutional neural network(DCNN) to effectively use the expanded data to improve the ability to recognize defective images. Finally, through comparative experiments, it is shown that this expanded data has an important role in improving the training of the defect detection network.

关 键 词：CyclyGANs 数据扩充 缺陷识别 无损检测 缺陷数据生成 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]