Infrared Image Detection Algorithm of electrical equipment Based on Improved YOLOv8

There are difficulties in the infrared image detection of substation equipment, such as large equipment size span, diverse types, and low detection accuracy of small-sized equipment. So A detection algorithm based on the YOLOv8 model, ET-YOLO (ELA-HSFPN-TADDH), is proposed to address those challenging problems. ELA-HSFPN, the Enhanced Lightweight Attention High Selective Feature Pyramid, is adopted to improve the model's feature fusion ability, fuse high-level and low-level feature maps, and obtain rich detailed information. Then the Lightweight Task Alignment Dynamic Detection Head (TADDH) is employed to combine scale, space, and task attention mechanisms to reduce model size and improve detection efficiency. For the ET-YOLO algorithm, the evaluating indicator [email protected] reaches 90.9% and is better than other tested models. The parameter and computational complexity of ET-YOLO were reduced by 31.25% and 17.43% respectively compared to the YOLOv8 model. The experimental results show that the ET-YOLO algorithm can meet the requirements of accurate real-time identification of substation equipment and lay the foundation for subsequent substation equipment fault diagnosis.