A review on image classification of remote sensing using deep learning

摘要

Deep learning is the state-of-the-art of machine learning. Previous literature demonstrates that deep learning gains the excellent performance in image classification of remote sensing. To begin with, data sources of remote sensing and current classification methods are briefly introduced. Then, common data sets and typical models of deep learning are presented, including deep belief network, convolutional neural network, stacked auto encoder. Furthermore, optimal configuration of these methods of deep learning is summarized according to the overall accuracy and Kappa coefficient. Finally, the existing problems and future work of satellite images classification by deep learning are pointed out. The review shows that deep learning is promised to be dominant method of image classification of remote sensing.