The concentric optical camera based on the optical fiber relay image can meet the needs of the miniaturization, lightweight, and large field-of-view detection of the space target detection system. However, due to the ...
详细信息
The concentric optical camera based on the optical fiber relay image can meet the needs of the miniaturization, lightweight, and large field-of-view detection of the space target detection system. However, due to the inevitable thermal stretching and torque processes involved in the production of fiber-optic panels (FOPs), defects of unknown shape, size, and location are formed, increasing the difficulty of space target information acquisition and causing center-of-mass localization errors. To efficiently and accurately detect and locate defects in FOPs, we propose an improved defect detection and localization method for FOPs based on discrete wavelet and multiobjective morphological optimization by using the gray value of the image sensor. We compare this method with the Canny edge detection method and the fuzzy C-means (FCM) clustering method and find that the method has higher detection accuracy and stronger robustness in an environment with a large number of dark filament interferences, and the detection results are not affected by the various image sizes of the FOPs. The method also maintains a high degree of consistency in detection metrics, with the F_Score detection accuracy of 85%. This research provides a reliable solution for FOP defect detection and localization, lays the foundation for image quality improvement of concentric optical cameras based on fiber-optic relay imaging, and contributes to space target detection and localization.
PurposeThis paper aims to present a data-processing methodology combining kernel change detection (KCD) and efficient global optimization algorithms for solving inverse problem in eddy current non-destructive testing....
详细信息
PurposeThis paper aims to present a data-processing methodology combining kernel change detection (KCD) and efficient global optimization algorithms for solving inverse problem in eddy current non-destructive testing. The main purpose is to reduce the computation cost of eddy current data inversion, which is essentially because of the heavy forward modelling with finite element method and the non-linearity of the parameter estimation problem. Design/methodology/approachThe KCD algorithm is adapted and applied to detect damaged parts in an inspected conductive tube using probe impedance signal. The localization step allows in reducing the number of measurement data that will be processed for estimating the flaw characteristics using a global optimization algorithm (efficient global optimization). Actually, the minimized objective function is calculated from data related to defectdetection indexes provided by KCD. FindingsSimulation results show the efficiency of the proposed methodology in terms of defect detection and localization;a significant reduction of computing time is obtained in the step of defect characterization. Originality/valueThis study is the first of its kind that combines a change detection method (KCD) with a global optimization algorithm (efficient global optimization) for defectdetection and characterization. To show that such approach allows to reduce the numerical cost of ECT data inversion.
Industrial production processes present challenges in collecting and annotating defect samples, primarily due to cost constraints and the inability to comprehensively cover all defect categories. To address this issue...
详细信息
Industrial production processes present challenges in collecting and annotating defect samples, primarily due to cost constraints and the inability to comprehensively cover all defect categories. To address this issue, recent research has focused on modeling normal sample features. This article proposes DualFlow, an unsupervised anomaly detection and localization algorithm based on normalizing flow. DualFlow uses a dual-branch architecture to effectively balance the detection and localization capabilities of existing algorithms. To handle features of different scales, we propose a compact and efficient module called the gated multiscale feature fusion (GMFF) module. In addition, DualFlow incorporates a variance stability loss to exploit the inherent stability of normal sample features, resulting in a significant reduction in false-positive instances. DualFlow is designed for end-to-end training and inference, greatly enhancing defect detection and localization capabilities. DualFlow achieved 99.33% image-level area under the receiver operating characteristic (AUROC) and 98.05% pixel-level AUROC on the MVTec AD dataset. Furthermore, DualFlow exhibits robust generalization performance, as confirmed by its evaluation on the more challenging MVTec LOCO AD, BTAD, VisA, and KolektorSDD2 datasets.
暂无评论