Improving Impervious Surface Estimation by Using Remote Sensed Imagery Combined With Open Street Map Points-of-Interest (POI) Data

Abstract

As a key indicator of urban environments, the accurate mapping of impervious surface is essential. With the availability of high spatial resolution remote sensing imagery, such as Sentinel-2, it is feasible to produce fine-scale impervious surface maps. For mapping high-resolution impervious surfaces, object-based image analysis (OBIA) classification method demonstrated its efficiency and accuracy by combining spectral information and spatial information. Although with some success, impervious surface estimation remains challenging because different land cover types share similar spectral information. With the emergence of affordable GPS-enabled devices (such as a smartphone) and web 2.0 development, more and more people are getting involved in sharing their locations with others or posting on the Internet. These volunteered geographical information (VGI) data provide a brand-new prospect for mapping urban impervious surfaces. This research proposes an optimized method for impervious surface mapping based on Sentinel-2 multispectral imagery and open street map (OSM) points-of-interest (POI) data. The proposed method was tested in Milwaukee county, US, and the results show that the overall accuracy of the proposed OBIA increases from 82.57% to 87.02% compared with the conventional OBIA. Thus, this study provides an effective means of combining OBIA and the relatively new VGI POIs data to extract impervious surface with higher spatial resolution.