摘要
定期对复合绝缘子的憎水性进行检查,及时更换严重老化的复合绝缘子,是保障电网安全稳定运行的关键。为实现无人机图像拍摄后复合绝缘子憎水性的准确判别,提出一种基于候选区域的快速卷积神经网络绝缘子憎水性分析方法,利用训练得到的Faster R-CNN深度神经网络模型精确定位复杂背景中的复合绝缘子伞裙,通过位置修正提取其中水迹信息区域,最后送入图像分类卷积神经网络完成憎水性等级的判定。测试结果表明,该方法在无人工参与的情况下,能在复杂背景图像中准确定位复合绝缘子伞裙水迹区域,憎水性等级判定结果达到了较高的准确率。
Regularly checking the hydrophobicity of composite insulators and replacing the seriously aged composite insulators in time are the key to ensure the safe and stable operation of the power grid. In order to accurately identify the hydrophobicity of composite insulator after the image taken by the drone, a method of hydrophobicity analysis for composite insulators based on Faster R-CNN neural network is proposed in this paper. A Faster R-CNN deep neural network model is trained to accurately locate the composite insulators umbrella skirt in complex background. The region of water trace information is extracted through the position correction, and the image classification convolution neural network is sent to determine the hydrophobicity level. The results show that the method can accurately locate the water area of composite insulators without artificial participation, and the result of hydrophobicity level determination is of high accuracy.
作者
张德钦
刘晓伟
刘源
江振钰
夏鹏
ZHANG Deqin;LIU Xiaowei;LIU Yuan;JIANG Zhenjue;XIA Peng(Guangxi Power Grid Company Limited Beihai Power Supply Bureau,Beihai 536000,China;Wuhan University,Wuhan,430072,China)
出处
《智慧电力》
北大核心
2019年第4期104-109,117,共7页
Smart Power
基金
国家自然科学基金资助项目(41671419)~~
