Agricultural Land-Use Classification on Satellite Data Using Machine Learning

Abstract

Background: The utilization of satellite images has become increasingly popular for detecting land usage, with a focus on agricultural land classification in recent years due to the significant decline in the number of bees. Objectives: This paper seeks to address these challenges by applying several machine learning algorithms on multi-spectral satellite data from Sentinel-2 to derive accurate land classification models. Methods/Approach: Specifically, we use five bands: Red, Green, Blue, NIR, and NDVI to build three models, namely Random Forest (RF), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). Results: Our results show that on collected satellite data, the CNN model outperforms the other algorithms, with an accuracy score of 0.82, F1-score of 0.72, and AUC score of 0.94, followed by the RF and LSTM models. Conclusions: This highlights the importance of utilizing advanced machine learning techniques, particularly CNNs, in accurately classifying agricultural land use changes.