Agricultural Land Classification Using Svm Classifier

Authors

  • Kalathur.Rakesh, S Dhamodaran, Tummalapalli Ganesh, Viji Amutha Mary A, P. Jeyanthi

Abstract

Agriculture research has gained traction in recent years and is exhibiting signals of significant expansion. The most recent entrant to the scene is bringing convenience to agriculture using various computational technologies. We used LAND satellite pictures to implement this research, which include images of FOREST, Agricultural land, urban area, and range land are all included. For Sentinel-2's Multispectral Imager (MSI), only a few studies have been conducted. In MSI, classifier performance on the same remote sensing pictures with varying training sample sizes were tested. RF, KNN, and SVM classifiers were analysed and evaluated using Sentinel-2 images. It was found that the Red River Delta in Vietnam was researched using 14 various training sample sizes, including both balanced and imbalanced ones. For the categorization results, an OA of 90% to 95% was recorded in this trial. Training sample size was less of an issue for SVM than other methods. Using a large enough amount of data to train each classifier resulted in excellent accuracy (i.e., more than 750 pixels per class, or around 0.25 percent of the total research area). Even with data that was both unbalanced and unevenly distributed, these remarkable outcomes were nonetheless achieved.

Downloads

Published

2022-07-23