基于ML S2OCELM的输电线异物检测方法研究
    点此下载全文
引用本文:陈玉权,王红星,沈杰,张星炜,黄郑,李学钧.基于ML S2OCELM的输电线异物检测方法研究[J].南京邮电大学学报:自然科学版,2020,40(3):89~96
摘要点击次数: 69
全文下载次数: 38
作者单位
陈玉权 江苏方天电力技术有限公司江苏南京210036 
王红星 江苏方天电力技术有限公司江苏南京210036 
沈杰 江苏方天电力技术有限公司江苏南京210036 
张星炜 江苏方天电力技术有限公司江苏南京210036 
黄郑 江苏方天电力技术有限公司江苏南京210036 
李学钧 江苏方天电力技术有限公司江苏南京210036 
基金项目:江苏方天电力技术有限公司科技项目(KJ201915)资助项目
中文摘要:输电线路上的异物可以视为输电系统的一种潜在风险,其不仅会影响输电线路的正常供电,还会对线下的行人与车辆产生较大的威胁。基于无人机航拍输电线路图像的异物检测已经得到初步的研究,但是依然存在算法复杂度高,特征表达能力弱与需要大量人工标注等问题。因此,文中提出了一种基于多层半监督单类极限学习机(ML S2OCELM)的输电线异物检测方法,具有特征表达能力强,运算速度快、需求标注少等优点。实验证明,文中提出的方法可以在极少的正标注情况下,高效率地筛选出存在异物的航拍图像,以帮助工作人员确定出现异物的输电线位置。
中文关键词:输电线  异物检测  极限学习机  无人机  半监督
 
ML S2OCELM based foreign body detection on transmission line
Abstract:The foreign body in the transmission line can be regarded as a potential risk of the transmission system.The risk affects the normal power supply of the transmission line and poses a great threat to the under pedestrians and vehicles.The foreign body detection based on UAV aerial transmission line image has been preliminarily studied, but there are still some problems, such as high algorithm complexity, weak feature expression ability and large amount of manual labelling. Therefore, a foreign body detection method is proposed for the power transmission line based on multi layer semi supervised single class extreme learning machine (ML S2OCELM).The method has the advantages of strong feature expression ability, fast operation speed and less requirement labeling. Experimental results show that the method can efficiently screen out aerial images with foreign bodies under the condition of few positive labels, so as to determining the location of the power line with foreign bodies.
keywords:transmission lines  foreign body detection  extreme learning machine  unmanned aeriaa vehicle (UAV)  semi supervision
查看全文  查看/发表评论  下载PDF阅读器

你是第3044563访问者
版权所有《南京邮电大学学报(自然科学版)》编辑部
Tel:86-25-85866913 E-mail:xb@njupt.edu.cn
技术支持:本系统由北京勤云科技发展有限公司设计