Deep learning-based automated fault diagnosis in transmission lines using UAV imagery

Reliable inspection of high-voltage transmission lines is essential for ensuring power grid stability and preventing unexpected failures. Conventional inspection approaches rely on manual patrols or rule-based image analysis, which are time-consuming, costly, and often unreliable under complex environmental conditions. To address these limitations, this paper proposes an adaptive deep learning based framework, termed AIWO–YOLOv9, for automated transmission line fault diagnosis using unmanned aerial vehicle (UAV) imagery. The proposed approach integrates the YOLOv9 object detection architecture with Adaptive Invasive Weed Optimization (AIWO) to dynamically tune key hyperparameters, improving detection accuracy and training stability while preserving real-time performance. The framework is evaluated on a UAV-based transmission line inspection dataset containing multiple fault categories, including damaged insulators, conductor defects, and vegetation encroachment. Extensive experiments show that the proposed AIWO–YOLOv9 model achieves a precision of 95.1%, recall of 93.4%, and an F1-score of 94.2%, outperforming state-of-the-art comparison models. In terms of localization accuracy, the proposed framework attains a mean Average Precision of 96.6% at IoU 0.5 (mAP@0.5) and 73.8% at IoU thresholds from 0.5 to 0.95 (mAP@0.5:0.95). Furthermore, the model maintains a real-time inference speed of 61 frames per second, demonstrating its suitability for practical UAV-based inspection applications. The results confirm that AIWO–YOLOv9 provides a robust balance between detection accuracy, localization precision, and computational efficiency. Owing to its scalability and adaptive optimization capability, the proposed framework offers a reliable solution for intelligent transmission line monitoring and condition-based maintenance in modern power grid systems.