Aiming at the current problem of low resolution and contrast of infrared images of power equipment

which causes the decrease of power equipment detection accuracy

a target detection method based on super-resolution reconstruction and DE-YOLOv11 is proposed. First

a super-resolution network is constructed

integrating information distillation and large kernel attention mechanism to extract and enhance the multi-level features of the output feature map. Then

this paper introduces a dual feature fusion module in YOLOv11 to take into account both local and high-frequency information extraction

and the detection head is improved using a shared convolutional layer to improve the model’s multi-scale sensing ability while maintaining the lightweight. Subsequently

on the basis of the reconstructed image

this paper constructs a detection network to effectively recognize the key features of power equipment. Finally

experimental results demonstrate that the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) of MAN on the infrared image dataset of power equipment are 36.312dB and 0.9593

respectively. Meanwhile

the DE-YOLOv11 model achieves a precision of 93.1%

an mAP50 of 95.5%

and a recall of 90.8% on the reconstructed dataset

which obviously outperforms several other recognition algorithms. The proposed method effectively enhances the accuracy of power equipment infrared image recognition

providing robust support for improving the precision of power equipment fault diagnosis.