A Comparative Analysis of Different Land-use and Land-cover Classifiers using Hyperspectral Data

In the past years, various efforts have been made to extract critical information for land-use and land-cover (LULC) areas using hyperspectral imagery which was not possible with the help of multispectral imagery. Remote sensing plays an important role in providing information over large areas and in monitoring the various changes over LULC. Classification is one of the main methods which is used for the detection of changes over the earth's surface. The evaluation of different classifiers using hyperspectral imaging is essential be performed. Therefore, the main objectives of current research work are to analyze and implement the various supervised classifiers such as support vector machine (SVM), neural networks (NN), maximum likelihood classifier (MLC), and K-means as an unsupervised classifier. These classifiers have been implemented and evaluated using a hyperspectral dataset over a part of Haryana and Uttar Pradesh states, India. The results have shown that NN (89.20%) algorithm has achieved higher accuracy than other algorithms and is useful in the mapping of changes over LULC using hyperspectral imagery. This study is useful in many applications such as monitoring and mapping seasonal variations and natural resources.