Application of Machine Learning Algorithms on Temporal Data for Vegetation Indices in Crop Classification

This paper therefore aims at examining the ability of machine learning models on the temporal satellite data for crop classification with the insert of the spectral indices; NDVI, EVI and SAVI. Accurate crop identification was achieved by assessing three models; namely Random Forest, Support Vector Machine (SVM) and Long Short Term Memory (LSTM) networks developed using multi-temporal satellite images. The experiment reveals that the LSTM model, which captures temporal behavior of the input data, yielded the highest classification score, 91%, compared to both Random Forest and the SVM model. The results show that the temporal data are useful in differentiation between crops with similar spectra but different phenology. The results of this research confirm that timely data coming together with state-of-the-art machine learning algorithms can further enhance the accuracy of crop classification, thus helping to enhance the efficiency of precision agriculture. The work to be done in the future will be to include other sources of climate data such as meteor and soil to improve model stability and performance.