Research on Infrared Image Recognition Algorithm for Power Equipment Based on Improved YOLOv8n

To solve the problem of low recognition precision and high missed detection rate caused by complex backgrounds and blurred target edges in infrared images of power equipment, An improved YOLOv8n (You Only Look Once Version 8 nano)-based power equipment infrared image recognition algorithm is proposed. In the backbone network of this algorithm, a Global Edge Information Transfer (GEIT) module is designed. An edge detection operator is combined with the C2f module to form a C2f module for edge information generation. Through these two modules, edge information and spatial information are fused, thereby improving the recognition accuracy of different power equipment. Meanwhile, the Context Anchor Attention (CAA) mechanism and WIoUv2 loss function are used to optimize the localization precision in complex scenarios and reduce the missed detection rate. Experimental results show that the improved algorithm achieves a precision (P) of 89.7%, a recall (R) of 76.9% on the self-built power equipment dataset, with mAP@50 and mAP@50:95 reaching 85.5% and 48.8% respectively. Compared with the benchmark YOLOv8n algorithm, the precision is increased by 4.3%, the recall is increased by 0.3%, and the mAP@50 is increased by 1.6%.