IMPROVING LULC CLASSIFICATION FROM SATELLITE IMAGERY USING DEEP LEARNING – EUROSAT DATASET

Abstract. Machine learning (ML) has proven useful for a very large number of applications in several domains. It realized remarkable growth remote-sensing image analysis over the past few years. Deep Learning (DL) subset machine were applied this work to achieve better classification Land Use Cover (LULC) satellite imagery using Convolutional Neural Networks (CNNs). EuroSAT benchmarking data set is used as training which uses Sentinel-2 images. provides images with 13 spectral feature bands, but surprisingly little attention been paid these features deep models. The majority focused only on RGB due high availability models computer vision. While gives an accuracy 96.83% CNN, we are presenting two approaches improve performance In first approach, extracted from bands instead leads 98.78%. second approach addition calculated indices LULC like Blue Ratio (BR), Vegetation index based Red Edge (VIRE) and Normalized Near Infrared (NNIR), etc. 99.58%.