Research on surface damage identification method of transmission towers based on improved single-stage detection algorithm

Abstract

As critical infrastructure in power transmission systems, the structural condition of transmission towers directly affects grid stability and power supply safety. Damages such as cracks, corrosion, and missing bolts may develop over time due to environmental factors and external loads. Although UAV-based inspection has become a mainstream data acquisition method, accurate damage identification remains hindered by complex backgrounds and insufficient small-target detection. To address these challenges, this study proposes an improved object detection algorithm named YOLOv11-TA-Convformer, incorporating a multi-attention mechanism. Firstly, the Convformer module was introduced to replace the original C3K2 module in YOLOv11 backbone network to enhance feature extraction capabilities, while the Triplet Attention (TA) mechanism was integrated to improve focus on critical features. Secondly, a dataset comprising four types of transmission tower damage, including pier cracks, steel corrosion, missing bolts, and structural steel cracks, was established using data augmentation techniques. The impacts of different optimizers, including SGD, Adam, and RMSProp, along with various learning rate strategies, were analyzed during the training phase. Finally, comparisons were made with alternative attention mechanisms such as SENet and CBAM, alongside contemporary state-of-the-art detection algorithms. Comparative results demonstrate that SGD achieves the fastest convergence and the lowest final loss, and is therefore selected as the optimizer. Experimental results demonstrate that the YOLOv11-TA-Convformer model achieved a mean average precision (mAP) of 0.873, surpassing all baseline models and variants. Engineering applications validated the effectiveness, improving inspection efficiency by 2.8 times and reducing annual maintenance costs by 17.3%. This research provides an accurate solution for identifying transmission tower damages and offers reliable technical support for intelligent grid maintenance, paving the way for wider engineering deployment.