Satellite Image Classification and Change Detection: A Case Study of Almora Town, Uttarakhand, India

Abstract

Land use/ land cover classification using satellite imagery is gaining abundant attention of researchers for extracting information from geospatial data. Various techniques are popular for the information extraction process from images to monitor a large geographical area and a comparison between these classifiers is required to select the appropriate one for the given application. The current study compares the four different classifiers namely Maximum Likelihood, Minimum Distance, Mahalanobis Distance, and Parallelepiped taking Almora town as the test area, which is located at Almora district of Uttarakhand province. Landsat imagery from two different years, i.e. 1999 and 2020, obtained from USGS Earth Explorer portal of geospatial datasets, has been used for the case study. First, the 2020 image of the study area has been classified using aforementioned four classifiers, and the accuracy of the four classifiers evaluated. Based on the accuracy level, the 1999 and 2020 images further classified with the classifier having highest accuracy rate to detect the changes in the given two decades.