Urban footprint detection from night light, optical and SAR imageries: A comparison study


Baydogan E., SARP G.

REMOTE SENSING APPLICATIONS-SOCIETY AND ENVIRONMENT, vol.27, 2022 (Peer-Reviewed Journal) identifier identifier

  • Publication Type: Article / Article
  • Volume: 27
  • Publication Date: 2022
  • Doi Number: 10.1016/j.rsase.2022.100775
  • Journal Name: REMOTE SENSING APPLICATIONS-SOCIETY AND ENVIRONMENT
  • Journal Indexes: Emerging Sources Citation Index, Scopus
  • Keywords: Urban footprint, Speckle divergence approach, Human settlement index (HSI), Support vector machines (SVM), Istanbul, MAPPING URBANIZATION DYNAMICS, SUPPORT VECTOR MACHINES, CITY LIGHTS, FUSION APPROACH, AREAS, SCALE

Abstract

The rise of large cities and their increasing spatial effects cause strong imbalances between the city and its hinterland. In such cities, the need for effective and timely handling of urban planning and management processes has been strongly felt in recent years. This situation has led to an increasing demand for temporal data with high spatial accuracy. Today, with the existence of Spaceborne Earth observations that meet this demand, it has become possible to follow the spatial distribution of urban footprints and their development over time in constantly changing big cities. In this study, urban footprints were determined from optical Landsat 8 Operational Land Imager (OLI) images using Support Vector Machines (SVMs) image classification, from Luojia 1-01 (LJ101) night light data using the Human Settlement Index (HSI), and from Sentinel-1A Synthetic Aperture Radar (SAR) data using the speckle divergence approach. According to the results, the general accuracy of urban footprints obtained by the HSI method from night images of the Lj1-01 satellite is 92%. The general accuracy of urban footprint obtained with SVM classification from optical images is 86%. The general accuracy of urban areas obtained with the Speckle Divergence approach for Vertical-Vertical (VV) and Horizontal-Vertical (VH) polarizations are 83% and 81%, respectively. The accuracies obtained with the analyzes revealed that the data and methods used in the study can be effectively used in determining the urban footprints.