A sample expansion method and system for power distribution network inspection defect identification
High-quality distribution network inspection images are generated by gradient field-guided fusion and group normalization techniques, which solves the problem of sample scarcity in distribution network inspection, improves sample diversity and network stability, and supports the application of deep learning algorithms.
CN117253107BActive Publication Date: 2025-12-12STATE GRID INTELLIGENCE TECHNOLOGY CO LTD
Patent Information
- Application Number
- CN202311226875.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-21
- Publication Date
- 2025-12-12
- Estimated Expiration
- 2043-09-21
Smart Images
Figure CN117253107B_ABST
Abstract
The present application relates to the technical field of power inspection, and provides a sample expansion method and system for distribution network inspection defect identification, comprising: acquiring a distribution network inspection image and a defect description; adding noise to the distribution network inspection image to obtain a temporary image matrix; converting the defect description into a text vector, and fusing the text vector and the temporary image matrix using gradient field guided fusion; based on the fused result, using a neural network to predict noise; and converting the temporary image matrix after subtracting the predicted noise into a generated image. The method ensures the generation of high-quality distribution network inspection defect images, and realizes the expansion and processing of small sample distribution network inspection images.
Need to check novelty before this filing date? Find Prior Art
Citation Information
Patent Citations
Magnetic core surface defect detection method based on deep learning
CN113888485A
Image redrawing model training method, image redrawing method and device
CN116664719A