To tackle challenges in identifying small targets

weak feature representation

and significant background interference in gas-insulated switchgear (GIS) defect detection

this paper proposes a high-precision defect detection model that integrates multi-layer feature interaction and an attention modulation mechanism. In the feature fusion stage

the model introduces the ASF (attention scale sequence fusion)-YOLO (You only look once) model structure to enhance the expression consistency and detail perception capability of multi-level features through spatial and scale feature interaction. In feature extraction

a focal modulation module is embedded to achieve adaptive aggregation of global and local features through the focal modulation mechanism

thereby improving the sensitivity to small defects in complex backgrounds. Meanwhile

the backbone network integrates the focused linear attention (FLA) mechanism to optimize collaborative modeling of channel and spatial features and strengthen detail capture in key defect regions. Experimental results show that the proposed model significantly improves both detection accuracy and feature discrimination capability

providing an efficient solution for intelligent defect detection of power equipment.