CROP CLASSIFICATION USING A COMBINATION OF SPECTRAL INDICES FROM SPATIOTEMPORAL MULTISPECTRAL IMAGERY AND MACHINE LEARNING

This study aims to evaluate the potential of spatiotemporal multispectral (MSI) Sentinel-2A and Sentinel 2B for crop classification. These satellites have 13 bands covering the visible, near infrared and short-wave infrared (SWIR) wavelength regions, offer a vast number of vegetation indices. For generating crop maps, the three common approaches namely Kernel extreme learning machine (KELM), Random Forest (RF) and Support Vector Machine (SVM) were applied and compared. 82 Vegetation indices were added to improve classification accuracy. SVM yielded the highest performance to classify the six crop types includes: Beans, Beet, Grass, Maize, Potato and Wheat, achieving overall accuracies of 0.63, 0.82, 0.88 and 0.90. This study showed that combination of multispectral remote sensing data and machine learning algorithms had effective to crop type classification.