Anomaly discrimination technique combining small sample reliability improvement and combination strategy optimization

The safe operation of electrical equipment in substations is crucial. This article uses Mixup, Mosaic, and stuffing augmentation to expand data and proposes a known anomaly detection method based on combination strategy, significantly improving the detection accuracy of known equipment defects. A method for identifying unknown changes based on twin networks has been proposed, achieving precise analysis of unknown changes. This article's technology can achieve high robustness anomaly discrimination by combining known anomaly recognition and unknown difference analysis in substation inspection images, replacing the work mode of on-site staff inspections and reducing their workload. Not only does it reduce labor costs, but it also improves operational efficiency.