13 January 2022 in IEEE Transactions on Industrial Informatics
Unsupervised anomaly detection in real industrial scenarios is challenging since the small amount of defect-free images contain limited discriminative information, and anomaly defects are unpredictable. Although nowadays image reconstruction-based methods are widely being used in various anomaly detection applications, they cannot effectively learn semantic representation, which leads to imperfect reconstruction. In this article, anomaly detection is formulated as a joint problem of feature reconstruction and inpainting in the dual-siamese framework. The proposed approach forces the network to model the feature distribution from the normal area and capture the semantic context for discriminating normal and abnormal areas. It first uses a Siamese architecture to capture discriminative features of defect-free samples and its corresponding defective samples generated by the defect random generation module. A dense feature fusion module is then employed to obtain the dense feature representation of dual input. The second Siamese network is proposed to reconstruct and inpaint the dual-dense features of the previous stage. Compared to the existing methods that mostly employ single image reconstruction, it is beneficial to simultaneously reconstruct and inpaint the information of dense discriminative features. The experimental results on the MVTec AD datasets and some major real industrial datasets demonstrate that our method achieves state-of-the-art inspection accuracy.
7 June 2021 in IEEE Transactions on Instrumentation and Measurement
The detection of conductive particle images is an important part of the chip on glass (COG) process and can be used to ensure the performance of electrical connections. The segmentation of conductive particles is essential but a difficult task, since the scale and edge of the conductive particles on the chip and the imaging effect are different. In recent years, methods based on deep learning have become the representative method of image segmentation. However, the currently existing methods cannot fully consider the characteristics of conductive particles and have high model complexity. In this article, a multi-frequency feature learning-based convolutional neural network (CNN) is proposed. The entire network structure consists of a basic U-Net module and multi-frequency module (MFM), which are used to enhance multi-frequency feature fusion of conductive particles and accelerate network training. At the same time, for the feature of particle shape, an active contour without edge (ACWE) loss function is designed to extract the fine contour feature of particles. Experimental results on three datasets show the superiority of the proposed method over the major existing mainstream methods with respect to the three performance indicators: recall, precision, and Intersection-over-Union (IoU).