RS-YOLOv8: An Insulator Defect Detection Algorithm with Feature Extraction and Fusion Optimization

Due to the complex background of insulator image and various defect types, the existing deep learning-based object detection algorithms are difficult to achieve accurate insulator defect diagnosis. Therefore, this study proposes a novel insulator defect diagnosis algorithm RS-YOLOv8 based on feature extraction and fusion optimization. First, we introduce the Dilation-wise Residual (DWR) and Dilated Reparameterization (DRP) modules to improve the C2f structure of YOLOv8, proposing a new feature extraction and fusion module called RepDRC2f. This enhancement boosts the multi-scale feature extraction and multi-level feature fusion capabilities of YOLOv8. Secondly, by modifying the detection head, we propose a new SEHead structure, which further strengthens the capability to capture detailed edge features, effectively improving the accuracy of locating insulator defect areas. Finally, experimental results show that RS-YOLOv8 achieves a precision of 95.9 %, a recall of 86.8 %, and an mAP@ 0.5 of 93.8 %. Compared to YOLOv8, the precision, recall, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{mAP} {@} 0.5$</tex> are improved by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$4.2 \%, 2.3 \%$</tex>, and 4.7 %, respectively. This significantly enhances the accuracy of insulator defect detection.