Land Use and Cover Mapping Using SVM and MLC Classifiers: A Case Study of Aurangabad City, Maharashtra, India

Abstract

The fast developing cities like Aurangabad need an effective analysis of Land Use Land Cover. In this paper we examine the use of MLC and SVM for mapping the LULC from satellite Imagery. This paper investigates the accuracy of Support Vector Machine SVM and Maximum Likelihood Classification MLC for multi-spectral images of Aurangabad city and Waluj area. The satellite images collected from IRS-1C LISS III and PlanetScope Imagery. Our objective is to produce LULC map for Aurangabad and Waluj and estimate the change that happened to each class of the land cover. The results show that the increase of buildings area was significant from the period of 2008 to 2018. The accuracy of MLC was 94.13% and 85.65% with kappa 0.92, 0.81 in 2008 and 2018 respectively, and for SVM the accuracy was 93% with kappa 0.91, 94.4% with kappa 0.92 in 2008 and 2018 respectively. It was noticed that both SVM and MLC can be used effectively on LULC data analysis.