Cold Region Insulator Fault Video Image Detection Method Based on Yolov4-Tiny Network with Channel Pruning
CN118537646BActive Publication Date: 2025-10-14YICHUN POWER SUPPLY COMPANY OF STATE GRID HEILONGJIANG ELECTRIC POWER COMPANY +2
Patent Information
- Application Number
- CN202410670833.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-28
- Publication Date
- 2025-10-14
- Estimated Expiration
- 2044-05-28
Smart Images
Figure CN118537646B_ABST
Abstract
A method for detecting video images of cold-region insulator faults based on the Yolov4‑Tiny network with channel pruning belongs to the field of computer vision technology. It solves the problems of slow convergence rate of traditional insulator fault detection model training, structural and parameter redundancy, and the inability to detect faults on insulator images but not on videos. The present invention first constructs a data set based on the cold-region insulator images, and sequentially performs Yolov4‑Tiny network basic training, sparse training, channel pruning and fine-tuning training to obtain the final fault detection model. Through channel pruning and model training under multiple loss functions, the model size is compressed and the detection speed of cold-region insulator fault videos and images is improved. The present invention is mainly used for cold-region insulator fault detection.
Need to check novelty before this filing date? Find Prior Art
Citation Information
Patent Citations
Insulator defect detection method based on improved YOLOv5 convolutional neural network
CN112819804A
Target detection network batch normalization layer pruning method and device, equipment and medium
CN115953652A