Tian Yanfeng, Ahmad Rodina Binti, Abdullah Nor Aniza Binti
Faculty of Computer Science and Information Technology, Universiti Malaya, Kuala Lumpur, Malaysia.
PLoS One. 2025 Feb 25;20(2):e0318225. doi: 10.1371/journal.pone.0318225. eCollection 2025.
With the increasing demand for electricity, the safety and stability of power grids become paramount, highlighting the critical need for effective maintenance and inspection. Insulators, integral to power grid maintenance as protective devices on outdoor high-altitude conductors, are often subject to suboptimal image quality during drone-based inspections due to adverse weather conditions such as rain, snow, fog, and the challenges posed by sunlight, high-speed movement, and long-distance imaging. To address these challenges and achieve a more accurate inspection system, this manuscript introduces an insulator defect detection algorithm tailored for the low-quality images collected by drone-based imaging systems. Utilizing a patch diffusion model, high-quality images are obtained, enhancing the precision of insulator defect detection methods. Furthermore, to improve detection accuracy, we introduce an optimized DETR method that incorporates a Spatial Information Interaction Module to further strengthen the characteristics of minor defects. Additionally, a special convergence network is employed to augment the detection capabilities of the DETR. Experimental results demonstrate that our proposed insulator detection technique has achieved a detection accuracy of 95.8%, significantly outperforming existing defect detection methods in complex environments. It overcomes the drawbacks of traditional methods by employing sophisticated computational models, leading to more efficient, economical, and secure maintenance and inspection of power grids.
随着电力需求的不断增加，电网的安全与稳定变得至关重要，凸显了有效维护和检查的迫切需求。绝缘子作为户外高空导体上的保护装置，是电网维护不可或缺的一部分，但在基于无人机的检查过程中，由于雨、雪、雾等恶劣天气条件以及阳光、高速移动和远距离成像带来的挑战，其图像质量往往不尽人意。为应对这些挑战并实现更精确的检查系统，本文介绍了一种针对基于无人机成像系统采集的低质量图像量身定制的绝缘子缺陷检测算法。利用补丁扩散模型获取高质量图像，提高了绝缘子缺陷检测方法的精度。此外，为提高检测准确性，我们引入了一种优化的DETR方法，该方法结合了空间信息交互模块，进一步强化了微小缺陷的特征。此外，还采用了一种特殊的收敛网络来增强DETR的检测能力。实验结果表明，我们提出的绝缘子检测技术实现了95.8%的检测准确率，在复杂环境中显著优于现有缺陷检测方法。它通过采用复杂的计算模型克服了传统方法的缺点，从而实现了更高效、经济和安全的电网维护与检查。
