Classification of High Resolution Remote Sensing Images using Deep Learning Techniques

High Resolution Satellite Images are widely used in many applications. Since such images are useful to provide more useful information about the details about the every regions around the world. In this work, transfer learning is used efficiently for the feature extraction from a pretrained Convolutional Neural Network(CNN) model which is used for training in the classification task. Using transfer learning the classification yielded a better accurate results. The experiments are carried out on two high resolution remote sensing satellite images such as UC Merced LandUse and SceneSat Datasets. The pre-trained CNN used here is VGG-16 which is trained on millions of Image-Net Dataset. The proposed method yielded a classification accuracy of 93% in UC Merced LandUse Dataset and in SceneSat Dataset it is about 84%. This proposed method yielded a better precision of 0.93 and 0.86 in UC Merced LandUse Dataset and in SceneSat Dataset respectively.