Application of Improved YOLOv8 for Infrared Image Detection of Substation Equipment

To address the challenges of detecting distant, small targets in infrared images of substation electrical equipment, this paper proposes an improved YOLOv8 detection method. The enhanced YOLOv8 model incorporates an Efficient Multi-scale Attention (EMA) attention mechanism and a dual-channel feature network based on BiFPN, which strengthens the model's feature extraction and fusion capabilities. The original Complete Intersection over Union(CIOU)loss function is replaced with Weighted Intersection over Union(WIOU) to reduce the model's false positive rate. These improvements to YOLOv8 result in an 8.3 % increase in average detection accuracy for substation equipment infrared images and a 5.2% improvement in the P-R comprehensive index.