Land-cover Classification and Mapping for Eastern Himalayan State Sikkim

TL;DR

This paper presents an improved k-means and ANN-based approach for land-cover classification in Sikkim, addressing challenges posed by rapid urbanization and demonstrating superior accuracy over traditional methods.

Contribution

It introduces an improvised k-means algorithm and applies ANN for land-cover mapping, showing improved performance over existing classifiers in the region.

Findings

Improved k-means outperforms existing methods like k-NN and maximum likelihood.

ANN classifier demonstrates fast processing and high recognition accuracy.

The proposed methods effectively handle rapid land-cover changes in Sikkim.

Abstract

Area of classifying satellite imagery has become a challenging task in current era where there is tremendous growth in settlement i.e. construction of buildings, roads, bridges, dam etc. This paper suggests an improvised k-means and Artificial Neural Network (ANN) classifier for land-cover mapping of Eastern Himalayan state Sikkim. The improvised k-means algorithm shows satisfactory results compared to existing methods that includes k-Nearest Neighbor and maximum likelihood classifier. The strength of the Artificial Neural Network (ANN) classifier lies in the fact that they are fast and have good recognition rate and it's capability of self-learning compared to other classification algorithms has made it widely accepted. Classifier based on ANN shows satisfactory and accurate result in comparison with the classical method.