摘要
为实现大范围的省级国土三调分类数据与用地用海分类数据间的转换,本文采用深度学习技术和人工目视解译方法,根据国土三调分类与国土空间规划用地用海分类对接情况表,对国土三调分类二级类与用地用海分类二级类进行差异对比。首先,分析地类映射关系,研究“一对一”“一对多”地类映射关系的规律与特征,制作“一对一”转换工具,对“一对一”地类进行批量转换;其次,结合遥感影像、地名地址数据、地理国情监测数据、互联网POI等参考数据,采用深度学习技术方法,开展“一对多”地类自动转换方法研究;最后,研究一套半自动的国土三调地类与用地用海地类的转换技术流程,对基于深度学习的自动转换精度进行分析和评价,验证所提方法的可靠性和准确率。研究成果可实现广东全省国土三调地类向用地用海二级类的批量转换,大大提升转换效率,取得良好的效果,所提出的方法流程具有较强的科学性和可操作性。
In order to realize a wide range of conversion between the classification data of the third provincial land survey and the classification data of land-use&sea-use,deep learning technology and manual visual interpretation method are adopted,the differences between the second class of the third national land survey classification and the second class of land-use&sea-use classification are compared according to the table of the connection between the third national land survey classification and the land-use&sea-use classification of territorial spatial planning in this paper.Firstly,this paper analyzes the mapping relationship of land classes,studies the rules and characteristics of"one-to-one"and"one-to-many"mapping relationship,makes"one-to-one"conversion tool and converts"one-to-one"land classes in batches.Secondly,combined with remote sensing image,geographical name and address data,geographical monitoring data,internet POI and other reference data,deep learning technology is used to carry out the research on"one-to-many"automatic conversion method of land classes.Finally,a set of semi-automatic third national land survey classification and land-use&sea-use classification conversion process is studied,the accuracy of deep learning automatic conversion is analyzed and evaluated,and the reliability and high accuracy of the proposed method are verified.Research achievement realize the batch conversion of land classification of Guangdong Province's third land survey adjustment to the second class of land-use&sea-use classification,greatly improve the conversion efficiency,and achieves good results.The proposed method and process have strong scientific and operability.
作者
伍素贞
郭舟
淳锦
WU Suzhen;GUO Zhou;CHUN Jin(Guangdong Geological Surveying and Mapping Institute,Guangzhou,Guangdong 510815,China;School of Surveying and Mapping Science and Technology,Sun Yat-sen University,Zhuhai,Guangdong 519080,China;Guangdong Land Resources Surveying and Mapping Institute,Guangzhou,Guangdong 510599,China)
出处
《测绘标准化》
2024年第1期66-73,共8页
Standardization of Surveying and Mapping
基金
广东省智慧自然资源综合感知服务国土三调及其年度变更调查数据处理服务采购项目(1210-2241YDZB5221)。
关键词
一对一
一对多
地类转换
深度学习
参考数据
one-to-one
one-to-many
land class conversion
deep learning
reference data
