Automated Extraction of Urban Impervious Area from Spectral-Based Digital Image Processing Techniques

Abstract

Urban area comprises a complex mix of diverse land cover types and materials; it is often difficult to separate these classes due to their heterogenic nature. Studying and monitoring urban areas and its environment are closely associated with the study of impervious surfaces, which are anthropogenic features through which water cannot infiltrate into the soil. In the present study, spectral indices were developed using spectral information from satellite remote sensing sensor. Several spectral indices like vegetation index, soil-adjusted vegetation index (SAVI) and normalized difference vegetation index (NDVI); water index, modified normalized difference water index (MNDWI); and urban indices, normalized difference built-up index (NDBI), built-up index (BUI) and index-based built-up index (IBI), were implemented in the study. The combination of various spectral indices can be used, and finally using NDBI, BUI and IBI, principal component analysis (PCA) was performed, the first component of which was classified through unsupervised classification through K-means algorithm to extract urban built-up impervious features. The methodology has the potential to identify and automatically extract urban impervious features from other land use-land cover classes and is established over the city of Kolkata (India). Maps showing effective classification of urban areas were developed. The approach is further successfully operated over a forested area in order to extract settlements within the forest patch that proves the transferability of the method and can be universally accepted.