Abstract: In the context of equipment inspection within the complex environment of substations, challenges arise due to limitations in lighting conditions and sensor characteristics. These factors can make it difficult to detect target equipment from images captured under a single light source. To address this issue, a lightweight detection algorithm for substation equipment based on visible and infrared image fusion is proposed. Initially, a precise matching technique utilizing contour direction angles is employed to register visible light and infrared images of power transformation equipment, thereby achieving accurate alignment of dual-source images. Subsequently, guided filtering combined with saliency detection is applied to fuse visible light and infrared images, generating a comprehensive dataset of fused power transformation equipment images. Finally, a lightweight detection network tailored for power transformation equipment is designed. The original YOLOv10n model is enhanced by incorporating LS-Head and SC-C2F modules, as well as a loss function based on normalized Wasserstein distance. Experimental results demonstrate that, compared to the original model, the improved model reduces the number of parameters and computational load by 33.5% and 29.2%, respectively. Furthermore, the average accuracy reaches 94.8%, with a frame rate of 70.31 frames per second. This approach ensures high detection accuracy while achieving model lightweighting.