Remote Sensing-Based Crop Estimation Using Machine Learning

Abstract

Accurate crop estimation is vital for economic stability and food security. This research addresses the limitations of single-sensor optical mapping by integrating Sentinel-2 multispectral data with Sentinel-1 Synthetic Aperture Radar (SAR). We propose a Decision Fusion framework using Random Forest (RF), Support Vector Machine (SVM), and XGBoost. A pixel-level majority voting ensemble was implemented to classify Cotton and Rice in Bahawalnagar. Results demonstrate that the fused model achieved a peak accuracy of 94%, significantly reducing the spectral confusion between rice and other vegetation. Area analysis identified 729 acres of Cotton and 290 acres of Rice, providing a robust blueprint for regional agricultural monitoring.