Automated Detection Method of Line Defects Based on Container Technology

To address the challenges of low detection accuracy and insufficient real-time performance in the inspection of small targets on power transmission lines, this paper proposes an automated detection method based on container technology. First, data augmentation strategies (object copy-paste, multi-view transformation, and simulated illumination interference) and optimized label assignment are employed to mitigate dataset scarcity and imbalanced distribution. Second, a multi-level feature fusion network is constructed by integrating an improved Atrous Spatial Pyramid Pooling (ASPP) module with deep feature upsampling techniques, effectively preserving fine-grained details of small targets in high-resolution images. Finally, a containerized microservice architecture (Docker + FastAPI) is adopted to achieve lightweight deployment of the algorithm, supporting multi-modal input and real-time inference. Experimental results demonstrate that the proposed method achieves an F1-score of 0.75 and a mean average precision (mAP) of 0.747 on a self-built dataset of 9,838 transmission line defect images, significantly outperforming traditional detection approaches. Ablation studies validate the effectiveness of the data augmentation and feature fusion modules, while the system maintains high robustness in complex field environments, providing a reliable solution for intelligent power line inspection.