Sentinel-1 for object-based delineation of built-up land within urban areas

Abstract

This work deals with the delineation of built-up land within urban areas from Sentinel-1 data using object-based image analysis. The produced layers allow differentiation between built-up and non-built-up area. Additionally a layer is produced, presenting different types of built-up densities. The results are visually compared with a standardized product of the Copernicus earth observation program, the Copernicus High Resolution Layer Imperviousness Degree. For evaluation of the accuracy, the European Settlement Map 2016 was chosen as a reference data set. Results from the built-up density analysis are visually compared with reference layer generated from open government data. The results reveal the suitability of Sentinel-1 data for the delineation of built-up land within urban areas. The quality of the produced layers (built-up land map and built-up density map) is comparable to standardized products that are based on data from optical sensors e.g. Copernicus High Resolution Layer Imperviousness Degree, European Settlement Map 2016 or high resolution building density maps respectively. The accuracy of the built-up land map (BULM) is equal (78.2%) to the one of the settlement layer produced by use of the ISODATA cluster algorithm [1].