摘要
为了探索开发建设项目施工扰动区域和人为水土流失快速监测方法,针对特高压输变电工程线路长、施工标段多、扰动区域分散、人为水土流失严重、监督管理困难等特点,以榆横—潍坊输变电工程区为研究对象,利用卷积神经网络算法和高分2号卫星遥感影像,研究快速自动识别特高压塔基扰动区域、准确提取扰动面积的方法,选择平原农作物区、山地林区、丘陵草地区、平原草地区的33个样点(塔基)进行识别和分类计算。结果表明:利用卷积神经网络算法和高分2号卫星遥感影像可以快速识别施工扰动区域(范围),并准确提取施工扰动面积,与目视解译结果基本一致,与扰动面积实测值相比,相对误差最大值为11.77%、最小值为1.20%。
In view of the characteristics of UHV transmission and transformation projects such as long lines,many construction sections, scattered disturbed areas,serious human-induced soil erosion and difficult supervision and management,in order to explore the disturbed areas of construction projects and consider the rapid monitoring methods of soil erosion,this paper took Yuheng-Weifang Transmission and Transformation Project area as the research area and used Convolutional Neural Network Algorithm and Gaofen-2 satellite remote sensing image to study the method of fast and automatic identification of disturbed areas of UHV tower foundation and accurate extraction of disturbed areas.33 sampling points (tower foundations) such as plain crop area,mountain forest area,hilly grassland and plain grassland were selected for identification and classification calculation.The results show that the Convolutional Neural Network Algorithm and Gaofen-2 remote sensing image can be used to quickly identify the construction disturbance area (range) and accurately extract the construction disturbance area.The results are basically consistent with the visual interpretation results.Compared with the measured values of disturbance area,the maximum relative error is 11.77% and the minimum value is 1.20%.
作者
洪倩
陈枫楠
梁冬
李晋
陈晓枫
戚国辉
HONG Qian;CHEN Fengnan;LIANG Dong;LI Jin;CHEN Xiaofeng;QI Guohui(State Grid Economic and Technological Research Institute Co.,Ltd.,Beijing 102209,China;Unisplendour Software System Corporation Limited,Beijing 100084,China)
出处
《人民黄河》
CAS
北大核心
2019年第8期99-102,共4页
Yellow River
基金
国家电网公司科技项目(B3440918K004)
关键词
开发建设项目
水土流失监测
特高压塔基
扰动面积
卷积神经网络算法
高分2号卫星
自动识别
development and construction projects
monitoring of soil erosion
UHV tower foundation
disturbed area
Convolutional NeuralNetwork Algorithm
Gaofen-2 Satellite
automatic recognition
