Classification Of Hyperspectral Images Using Deep Learning Architecture for Remote Sensing Applications

Land use and land cover plays a vital role in the natural resource management. It is used to map the environmental changes for ecosystem monitoring. Automatic mapping can bring numerous values for agriculture, urban management and forestry. The advancement of deep learning techniques produces remarkable results on image classification filed which has applied essentially in the field of remote sensing. This paper presents land cover classification based on deep learning technique. It consists of U-Net architecture with encoder and decoder phases. The encoder internments semantic representation by utilizing the convolution layers. The decoder part reinstates spatial information with the help of skip connection. The proposed method is compared with state-of-the-art methods and produce 81.23% of over land cover datasets.