MSFANet: A Robust Multi-Scale Feature Extraction and Task Alignment Network for Insulator Defect Detection in Complex Environments

Detecting defects in power transmission line insulators is crucial for ensuring the safe and stable operation of the power grid. Traditional manual inspection methods are inefficient, particularly under adverse weather conditions and in complex terrains. To address this issue, we propose MSFANet (Multi-Scale Feature Alignment Network), an advanced insulator defect detection algorithm based on YOLOv8n. We design the Focusing Diffusion Pyramid Network (FDPN) to enhance multiscale feature extraction through specialized convolution and fusion techniques. We also introduce the Task Dynamic Align Detection Head (TDADH), which improves coordination between classification and localization tasks through task alignment strategies and dynamic feature selection. Furthermore, the Skew Intersection over Union (SIoU) loss function is employed to better capture the similarity between predicted and ground truth bounding boxes. Compared to the original YOLOv8n model, the improved model significantly enhances detection accuracy and recall rate in complex backgrounds, with mAP50 increasing by 6.0%, precision by 6.7%, and recall rate by 8.3%, while reducing the number of parameters to 2.6 million. The proposed MSFANet model excels in detecting insulator defects, particularly in handling complex backgrounds and small targets, showcasing its potential for real-time and efficient detection in power transmission lines.