全球30m分辨率人造地表覆盖数据集(2010)

全球30m分辨率人造地表覆盖数据集（2010）是国家基础地理信息中心领衔的团队自2009-2014年研发的数据集。该数据集是全球30m分辨率土地覆盖数据集（2010）的一部分，是反映人类居住、工矿、商业、交通等用地类型和人类生活场所、操作空间和工程的载体的数据集。人造地表覆盖类型的空间分布客观地反映了人类社会的资源配置、产业布局和经济状况。本文论及的数据集依托全球地表覆盖遥感制图项目，包括9907景美国陆地资源卫星（Landsat）TM5、ETM多光谱影像数据和2640景环境减灾卫星（HJ-1）多光谱影像数据。在基于卫星遥感数据基础上的土地覆盖分类信息提取过程中，人造地表覆盖类型的识别和信息提取是最复杂的。其主要原因是人造地表覆盖多以复合像元、镶嵌交错、零碎分散等方式布局。全球30m分辨率人造地表数据集（2010）的研发是根据人造地表覆盖在卫星遥感影像上反映的光谱、形态特征的多样性和复杂性，综合利用现有的分类技术，将像元分类器、对象分割和知识化判别有机结合，采取了基于“像元-对象-知识”的POK（Pixel-Object-Knowledge）分类方法完成。

The Global 30m Resolution Artificial Surface Cover Dataset (2010) was developed by a team led by the National Geomatics Center of China between 2009 and 2014. This dataset is a component of the Global 30m Resolution Land Cover Dataset (2010), which reflects land use types such as human settlements, industrial and mining areas, commercial zones, and transportation facilities, as well as the carriers of human living spaces, operational areas, and engineering constructions. The spatial distribution of artificial surface cover types objectively reflects the resource allocation, industrial layout, and economic status of human society. The dataset discussed in this paper relies on the Global Land Cover Remote Sensing Mapping Project, and includes 9907 scenes of multi-spectral image data from US Landsat TM5 and ETM sensors, as well as 2640 scenes of multi-spectral image data from China's Environmental and Disaster Monitoring Satellite (HJ-1). During the process of extracting land cover classification information based on satellite remote sensing data, the identification and information extraction of artificial surface cover types are the most complex tasks. The main reason is that artificial surface covers are mostly distributed in the forms of mixed pixels, mosaic interlacing, and fragmented dispersion. The development of the Global 30m Resolution Artificial Surface Cover Dataset (2010) was carried out based on the diversity and complexity of the spectral and morphological characteristics of artificial surface covers as reflected in satellite remote sensing images. It comprehensively utilized existing classification technologies, organically integrated pixel classifiers, object segmentation, and knowledge-based discrimination, and adopted the POK (Pixel-Object-Knowledge) classification method built on the "Pixel-Object-Knowledge" framework.