摘要
随着机器学习和无人机技术的不断发展,无人机电力巡检逐步取代了传统的人工巡检方式。近年来对配电线路已有相当数量的图像在线监测系统或图像巡视系统,如利用无人机装载可见光摄像机等进行巡线。但是在巡视结束后,这些系统会获取大量的图像信息,而大量的图像依然是由人来检查,智能化程度不高,而且如果对这些数据采用工作人员主观判断而没有图像自动分析功能的话,极易出现误判或漏判的情况,难以准确发现配电设备的安全隐患,不能满足智能电网建设的需要。因此,提出采用图像处理技术和机器学习中的核密度估计算法对航拍图像分类,将正常的图片和有绝缘子污闪的图片分开,减轻工作人员的工作量,并能够更快的定位和发现问题,满足智能配电网建设的需要。
With the continuous development of machine learning and drone technology,the drone power inspection has gradually replaced the traditional manual inspection method.In recent years to the distribution line has a considerable number of image online monitoring system or image inspection system,such as the use of drones loaded with visible light camera to patrol the line.But after the inspection,these systems will obtain a large number of image information,but a large number of images are still inspected by people,the degree of intelligence is not high,and if these data use staff subjective judgment without image automatic analysis function,it is very easy to misjudge or omit the situation,it is difficult to accurately discover the distribution equipment security hidden danger,cannot meet the needs of the smart grid construction.In this context,this paper adopts image processing technology and the kernel density estimation algorithm in machine learning to classify the aerial images and separate the normal images from the images with insulator dirt flashes to reduce the workload of the staff and to locate and discover the problems faster to meet the needs of intelligent transmission and distribution construction.
作者
王照
胡日鹏
葛馨远
李品磊
WANG Zhao;HU Ripeng;GE Xinyuan;LI Pinlei(Baiyun Power Supply Bureau,China Southern Power Grid,Guangdong 510403,China)
出处
《电子技术（上海）》
2020年第9期58-61,共4页
Electronic Technology
基金
中国南方电网有限责任公司科技项目(GZHKJXM20180046)
关键词
控制技术
电力巡检
机器学习
无人机
control technology
power patrol
machine learning
drones
