Mapping the World Population One Building at a Time

[摘要] High resolution datasets of populationdensity which accurately map sparsely distributed humanpopulations do not exist at a global scale. Typically,population data is obtained using censuses and statisticalmodeling. More recently, methods using remotely-sensed datahave emerged, capable of effectively identifying urbanizedareas. Obtaining high accuracy in estimation of populationdistribution in rural areas remains a very challenging taskdue to the simultaneous requirements of sufficientsensitivity and resolution to detect very sparse populationsthrough remote sensing as well as reliable performance at aglobal scale. Here, the authors present a computer visionmethod based on machine learning to create population mapsfrom satellite imagery at a global scale, with a spatialsensitivity corresponding to individual buildings andsuitable for global deployment.

[发布日期] 2017-12-15 [发布机构] World Bank, Washington, DC

[学科分类] 社会科学、人文和艺术（综合）

[关键词] POPULATION DENSITY;POPULATION ESTIMATE;SATELLITE IMAGERY;POPULATION DISTRIBUTION