摘要
以地理国情监测高分辨率遥感影像及高精度地表覆盖分类产品为数据源,提出了一种面向遥感影像智能分类、基于位置匹配技术的全国尺度海量样本数据采集方法。根据数据源特征,研究了县域采集数量权重设置、坐标投影转换、栅格灰度重采样、无效样本数据过滤、地表覆盖分类码映射、样本数据命名标识、特定地表覆盖类型样本数据采集等关键技术,构建了位置匹配的遥感影像数据与分类标签数据组成的样本数据对,开发了样本数据自动采集软件。利用该方法,以县级行政区划为单元,实现了全国尺度海量样本数据采集。选取其中5个县域的成果,评估了方法的实用性及运算性能。研究表明:该方法提升了生产全国尺度海量样本数据的计算响应速度;采集的样本数据能够满足遥感影像智能分类对样本源高质量、大规模的需求,提升了遥感影像分类与预测的准确度。
Based on the data sources of high-resolution remote sensing images and high-precision land cover classification products collected in the Geographic National Conditions Monitoring Project of China,a nationwide massive sample data acquisition method is proposed for intelligent classification of remote sensing images by using location matching technology.According to the characteristics analysis of data sources,the key technologies such as quantitative weight setting for each county,coordinate’s projection conversion,raster grid’s gray resampling,invalid sample data’s filtering,land cover’s classification code conversion,sample data’s file identification,and specific types of land cover’s sample data acquisition are researched.And a sample data pair formed by remote sensing image and classification label data is constructed.Besides,sample data automatic acquisition software is independent developed.By using this whole approach,the national scale massive sample data has been achieved,which is gathered by each unit of county level administrative division.The results of 5 different counties are selected to evaluate the practicability and operational performance of the method.The results show that this method can improve the calculation response speed of massive sample data produced once in whole country,and the collected sample data can meet the demand of high quality and large-scale sample sources for intelligent classification of remote sensing images,and improves the accuracy of classification and prediction of remote sensing images.
作者
程滔
吴芸
郑新燕
杨刚
白驹
CHENG Tao;WU Yun;ZHENG Xinyan;YANG Gang;BAI Ju(National Geomatics Center of China,Beijing 100830,China;Sinomaps Press,Beijing 100045,China)
出处
《测绘通报》
CSCD
北大核心
2019年第10期56-60,共5页
Bulletin of Surveying and Mapping
基金
国家地理国情监测专项(19-30-02-5)
国家基础地理信息中心科技创新发展基金课题(2018-KJ-G01)
关键词
遥感
地表覆盖
智能分类
样本
位置匹配
地理国情监测
remote sensing
land cover
intelligent classification
sample
location matching
geographic national conditions monitoring
