A Review on Deep Learning Models for Satellite Imagery

Authors : Hasan Ersan YAĞCI, Abdullah ATÇILI, Sukru SEZER

Pages : 73-79

View : 30 | Download : 19

Publication Date : 2021-09-05

Article Type : Research Paper

Abstract :Object detection and image classification from remote sensing data are used in many different fields. It has been the subject of many studies in recent years. Research in this field has increased with the development of deep learning techniques and remote sensing data, which can be satellite images or unmanned aerial vehicles (UAV), providing high resolution spatial and spectral data. In this review, we survey modern deep learning techniques are trained on remote sensing data. Term remote sensing data is widely used for satellite imagery, however the term also refers to UAV collected data. It is chosen as a topic of the this review that 'how green the metropolitans?'. There are two approaches for this question. First one is the detection of green (vegetation) in all metropolitan and the other one is classification of green types. Convolutional neural networks (CNN), generative adversarial networks (GAN), and autoencoder (AE) were compared on tensorflow's UC Merced dataset.
Keywords : Deep Learning, Satellite Imagery, Remote Sensing