URBAN FEATURE EXTRACTION USING HYPER-SHARPENING

Abstract

Satellite data has proven beneficial in quick monitoring of large urban areas and its various features. Urban centres are heart of country and concentrates large population and development so there regular mapping is essential for various applications such as monitoring City’s growth, Urban climate, Hydrological changes, Pollution of water and air, Natural disasters, Socio-economic factors. So, satellite data and its products are actually used to make cities more liveable and advance towards sustainability. This study aims at monitoring some of those crucial urban parameters- Impervious surface area, Pervious surface area and Urban Green space in one of the largest and densely populated cities in the world, Delhi in India. It was done by creating a finer scale LULC generated from a high Spectral-Spatial resolution Hyperspectral Data from PRISMA sensor. The Hyperspectral data was hyper-sharped from 30 m to 5 m spatial resolution using CNMF algorithm. It was then classified into LULC classes Water, Vegetation, Buildings, Roads and Bare land using machine learning algorithm and extracted for the required Urban features for two areas of the city. The maps of Impervious-Pervious surface areas and Urban green space were also created. The Areas of these urban features was also measured. At last, the results were compared with validation samples and higher resolution datasets for the assessment of accuracy. The classification accuracy for this method for two areas in Delhi was observed as 87 and 92%. The areal accuracies for urban features were found in range of 51.4 to 95.3 %.