This website requires JavaScript.

A Lightweight Reconstruction Network for Surface Defect Inspection

Chao HuJian YaoWeijie WuWeibin QiuLiqiang Zhu
Dec 2022
摘要
Currently, most deep learning methods cannot solve the problem of scarcity ofindustrial product defect samples and significant differences incharacteristics. This paper proposes an unsupervised defect detection algorithmbased on a reconstruction network, which is realized using only a large numberof easily obtained defect-free sample data. The network includes two parts:image reconstruction and surface defect area detection. The reconstructionnetwork is designed through a fully convolutional autoencoder with alightweight structure. Only a small number of normal samples are used fortraining so that the reconstruction network can be A defect-free reconstructedimage is generated. A function combining structural loss and $\mathit{L}1$ lossis proposed as the loss function of the reconstruction network to solve theproblem of poor detection of irregular texture surface defects. Further, theresidual of the reconstructed image and the image to be tested is used as thepossible region of the defect, and conventional image operations can realizethe location of the fault. The unsupervised defect detection algorithm of theproposed reconstruction network is used on multiple defect image sample sets.Compared with other similar algorithms, the results show that the unsuperviseddefect detection algorithm of the reconstructed network has strong robustnessand accuracy.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答