Insulator Umbrella Disc Shedding Detection in Foggy Weather.

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Stable operation of transmission lines relies on detecting insulator defects like umbrella disc shedding.
• Foggy weather significantly hinders the accurate visual inspection and detection of these critical defects.

Purpose of the Study:

• To develop an accurate and robust method for detecting insulator umbrella disc shedding in foggy aerial images.
• To address the challenges of limited real-world foggy image data and the small target size of shedding defects.

Main Methods:

• A two-stage approach integrating a defogging algorithm with a novel small object detection model (FA-SSD).
• Defogging utilizes synthetic data for initial training and real data for fine-tuning.
• FA-SSD fuses shallow and deep features and incorporates an attention mechanism for enhanced small target detection.

Main Results:

• The proposed model achieved a defect detection accuracy of 0.925 and a recall of 0.841 in foggy conditions.
• Significant improvements of 5.9% in accuracy and 8.6% in recall were observed compared to the original model.
• The model demonstrates effective detection of small insulator umbrella disc shedding defects even in adverse weather.

Conclusions:

• The combined defogging and FA-SSD detection model provides a reliable solution for identifying insulator umbrella disc shedding in foggy environments.
• This approach enhances the safety and efficiency of power transmission line maintenance.
• The method shows strong potential for real-world application in power infrastructure inspection.