基于遥感解译的中国农房空间分布特征分析

中国农房分布研究是乡村振兴的基础性工作之一。本文利用Mask R-CNN深度学习框架,采用亚米级遥感影像,基于中国各省农房的人文地理特征,分省训练出适用于本地的农房遥感解译模型,共识别1484个县的1.54亿栋农房,总建筑面积达298.2亿m²,创建出以单体农房为基础的多尺度中国农房数据库。本文分别以县域、镇域和聚落为单元进行农房密度分析：县域单元的农房密度呈现“东密西疏”的分布特征,与胡焕庸线揭示的人口分布格局相符;镇域单元的农房密度沿三大地形阶梯自西向东递增,分布从集中走向平均,与农业区划的农业生产力、地形地貌特征高度耦合;聚落单元的农房分布则呈现“南密北疏”特征形态,与气候条件、地形地貌、人均耕地等相关联。中国农房数据库的建设将为“可计算的”乡村研究奠定基础。

Spatial distribution of rural building in China: Remote sensing interpretation and density analysis

The research on the distribution of rural buildings is one of the essential works of rural revitalization in China. Adopting Mask R-CNN deep learning framework and collecting sub-meter remote sensing image, this research successfully trains remote sensing interpretation models suitable for local rural buildings based on the human geographical characteristics in each province. A multi-scale database of Chinese rural building based on single rural building has been created, including 154 million agricultural houses in 1484 counties with a total construction area of 29.82 billion square meters. Based on the database, this paper analyzes the distribution characteristics of rural buildings at the county, town and settlement scales. At the county scale, the density of rural buildings showed a distribution pattern of 'dense in the east and sparse in the west', consistent with the population distribution pattern revealed by Hu Huanyong Line. At the town scale, the density of rural buildings increases from west to east with the descent of three terrain steps, and the spatial distribution tends to be balanced, corresponding to the agricultural productivities and topography in the agricultural zones. At the settlement scale, the distribution of rural buildings shows the pattern of 'dense in the south and sparse in the north', consistent to the climate condition, topography and arable land per capita. The construction of China's rural building spatial database could lay a solid foundation for the "computational" rural research in the future.