基于深度学习的乱占耕地建房疑似图斑自动提取方法研究  

Research on Automatic Extraction Method of Suspected Patches of Random Occupation of Farmland and Building Based on Deep Learning

作　　者：王丹[1] 张璐[1] 柴燕妮 WANG Dan;ZHANG Lu;CHAI Yanni(Jiangsu Geomatics Center,Nanjing 210013,China)

机构地区：[1]江苏省基础地理信息中心,江苏南京210013

出　　处：《测绘与空间地理信息》2022年第6期51-53,57,共4页Geomatics & Spatial Information Technology

基　　金：2020年度江苏省自然资源科技计划项目——面向重要自然资源变化监测的遥感影像深度学习解译技术研究(2020034)资助。

摘　　要：为实现农村乱占耕地建房问题“早发现、早制止、早查处”,本文提出一种基于深度学习的乱占耕地建房疑似图斑自动提取方法,利用高分辨率遥感影像解译模型,结合第三次国土调查成果耕地数据,快速识别乱占耕地建房疑似图斑。最后以江苏省某县级市为例开展实验,实验表明:该方法提取结果准确率高、用时少,可为耕地保护、督察执法提供重要支撑。In order to quickly find out the building houses at cultivated fields,this paper proposes an automatic extraction method of suspicious illegal building patches at cultivated fields based on deep learning.Based on the interpretation model of high-resolution remote sensing images,combined with the cultivated land data of the third land survey,we can quickly extract suspicious illegal building patches.Taking a county-level city of Jiangsu Province as an example,the experiment shows that this method has high accuracy and less time,and then supplemented by manual verification,which can provide important support for farmland protection,supervision and law enforcement.

关 键 词：深度学习 乱占耕地 U-Net 自动提取 

分 类 号：P208[天文地球—地图制图学与地理信息工程]