Mapping Urban Functional Areas Using Multisource Remote Sensing Images and Open Big Data.

The urban functional areas (UFAs) are the basic spatial units of a city, identifying their category information and spatial distribution is of great significance for studying urban spatial structure and formulating scientific and reasonable urban planning policies. However, traditional classification systems are still limited due to the data processing costs and time-consuming process. With the development of high-resolution satellite, the 2-D image has been able to identify more ground object information on a fine scale, but it lacks the application of 3-D characteristics and social attributes. At the same time, machine learning and deep learning have shown great utility in extracting features, and they should be applied more to the classification of urban functional areas to improve efficiency. To solve these problems, we propose an efficient and accurate framework for mapping UFA using multisource geospatial data. It can grasp the dynamic changes in the city more accurately, and provide a reference for the study of urban functional areas. An improved frequency density (IFD) model was proposed to improve the overall classification accuracy by 4.4%. Besides, the nighttime light (NTL) data from SDGSAT-1 satellite glimmer imagers are also used to classify UFA. By combining NTL and urban building height data, a sky view enhanced nightlight index (SVENI) was proposed, which can improve the overall classification accuracy of UFA by 5.8%. This study systematically clarifies the role of data sources, methods, and automatic integration models in the classification framework for UFA, which is of great significance for urban planning and sustainable development.