Image-based Process Monitoring via Adversarial Autoencoder with Applications to Rolling Defect Detection

Yan, Hao; Yeh, Huai-Ming; Sergin, Nurettin Dorukhan

doi:10.1109/coase.2019.8843313

Cited by 10 publications

(2 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar ideas based on multi-scale features are also used by this literature [37,38]. Yan et al [39] proposed an adversarial auto-encoder network to monitor defective regions in rolling production. The method combined the power of GAN and the variational auto-encoder, which could be served as a nonlinear dimension reduction technique.…”

Section: Related Workmentioning

confidence: 99%

Neural Subspace Learning for Surface Defect Detection

Liu

Chen

et al. 2022

Mathematics

View full text Add to dashboard Cite

Surface defect inspection is a key technique in industrial product assessments. Compared with other visual applications, industrial defect inspection suffers from a small sample problem and a lack of labeled data. Therefore, conventional deep-learning methods depending on huge supervised samples cannot be directly generalized to this task. To deal with the lack of labeled data, unsupervised subspace learning provides more clues for the task of defect inspection. However, conventional subspace learning methods focus on studying the linear subspace structure. In order to explore the nonlinear manifold structure, a novel neural subspace learning algorithm is proposed by substituting linear operators with nonlinear neural networks. The low-rank property of the latent space is approximated by limiting the dimensions of the encoded feature, and the sparse coding property is simulated by quantized autoencoding. To overcome the small sample problem, a novel data augmentation strategy called thin-plate-spline deformation is proposed. Compared with the rigid transformation methods used in previous literature, our strategy could generate more reliable training samples. Experiments on real-world datasets demonstrate that our method achieves state-of-the-art performance compared with unsupervised methods. More importantly, the proposed method is competitive and has a better generalization capability compared with supervised methods based on deep learning techniques.

show abstract

Section: Related Workmentioning

confidence: 99%

Neural Subspace Learning for Surface Defect Detection

Liu

Chen

et al. 2022

Mathematics

View full text Add to dashboard Cite

show abstract

“…This process is illustrated in Figure 2. An AAE model training on a large number of samples can provide excellent reconstruction and control of latent code (Creswell et al., 2017; Yan et al., 2019).…”

Section: Introductionmentioning

confidence: 99%

Spectrum Monitoring of Radio Digital Video Broadcasting Based on an Improved Generative Adversarial Network

et al. 2021

View full text Add to dashboard Cite

Soon after the first radio communications were invented in the late 19th century, radio interference problem arose. Spectrum monitoring helps managers plan and use spectrum resources more effectively, avoids incompatible use, and identifies harmful interference sources promptly (Lu et al., 2017;Papadias et al., 2020). The importance of anomaly detection is due to the fact that anomalies in data translate to significant, and often critical, actionable information in a wide variety of application domains (Chandola et al., 2009;Zhou et al., 2021). For example, anomalies in the radio spectrum could suggest that the interfering emitters, malfunctioning equipment, or malicious users within their licensed bands is occurring in real scenarios (O'Shea et al., 2016). Many radio services such as mobile communication, FM broadcasting and radio digital video broadcasting (DVB), need to deploy a large amount of base stations and radio transmitting stations to improve wireless coverage to achieve their goals. Once the radio equipment is interfered and attacked, the received signal will contain anomalies and the routine work will be interrupted. To make matters worse, it is hard to detect abnormal behavior in these systems due to the high level of variation in exceptions or attacks (Feng et al., 2017). Therefore, it is necessary to study various spectrum monitoring technologies for detecting and defending against interference (Calvo-Palomino et al., 2020;Lu et al., 2017). Many methods have been proposed to monitor the radio spectrum (Ali & Hamouda, 2016;Qin & Li, 2020). And energy detection has an improving performance by increasing the sampling points under low signal-to-noise ratio (SNR) environment when the noise power is precisely known (determined SNR). However, when there is uncertainty in the noise power, it is no longer effective to manipulate the sampling points (Tandra & Sahai, 2008), and covariance detection has the ability to overcome this weakness (Jia et al., 2015), unfortunately, it is invalid for locating anomalies in the spectrum. Recently, some more competitive approaches have been proposed and insights progressed, such as detection based on convolutional neural network (Conn & Josyula, 2019), adversarial autoencoder (AAE)-based anomaly detector (Rajendran et al., 2019), hidden Markov model detection (Allahdadi et al., 2021).In most cases, data resources are highly imbalanced toward examples of normality, while lacking in examples of abnormality (Akcay et al., 2019), which makes anomaly detection more challenging. Since the pattern of normal signals in radio monitoring is known, anomaly methods based on the generative adversarial network (GAN) may be a decent solution. GAN originally proposed by Goodfellow and Pouget-Abadie (2014), has gained a lot of attention in the computer vision community due to their capability of data generation without explicitly modeling the probability density function. GAN is a powerful generative

show abstract