A Lightweight UAV-based Distribution Network Defect Detection Framework with Complementary Feature Enhancement and Depthwise Patch Compaction

UAV-based distribution network defect detection has emerged as a critical technology for ensuring the reliability and safety of modern distribution network systems. Still, existing frameworks often struggle with computational constraints on UAV platforms, suboptimal performance under small-sample conditions, and precision loss in multi-scale defect localization. In this study, we propose a lightweight detection framework tailored for UAV-based defect inspection. The framework integrates two key modules: (1) the Complementary Feature Enhancement (CFE) module, which fuses multi-level semantic and spatial features to enhance defect representation, and (2) the Depthwise Patch Compactor (DPC) module, which preserves contextual relationships through patch merging operations to improve scale robustness. Experimental results on a custom dataset of $\mathbf{1 0, 0 0 0}$ high-resolution images demonstrate that our method outperforms state-of-the-art methods across multiple metrics. This work provides a practical solution for edge-deployed UAV systems, enabling high-accuracy, low-latency defect monitoring in harsh environments.