TY - JOUR
T1 - Deep learning for remote sensing image classification
T2 - A survey
AU - Li, Ying
AU - Zhang, Haokui
AU - Xue, Xizhe
AU - Jiang, Yenan
AU - Shen, Qiang
N1 - Funding Information:
This research has received funding from the National Key Research and Development Program of China (Grant No. 2016YFB0502502), Foundation Project for Advanced Research Field (614023804016HK03002), and Shaanxi International Scientific and Technological Cooperation Project (2017KW-006).
Funding Information:
National Key Research and Development Program of China, Grant/Award Number: 2016YFB0502502; Foundation Project for Advanced Research Field, Grant/Award Number: 614023804016HK03002; Shaanxi International Scientific and Technological Cooperation Project, Grant/Award Number: 2017KW-006
Publisher Copyright:
© 2018 The Authors. WIREs Data Mining and Knowledge Discovery published by Wiley Periodicals, Inc.
PY - 2018/11/1
Y1 - 2018/11/1
N2 - Remote sensing (RS) image classification plays an important role in the earth observation technology using RS data, having been widely exploited in both military and civil fields. However, due to the characteristics of RS data such as high dimensionality and relatively small amounts of labeled samples available, performing RS image classification faces great scientific and practical challenges. In recent years, as new deep learning (DL) techniques emerge, approaches to RS image classification with DL have achieved significant breakthroughs, offering novel opportunities for the research and development of RS image classification. In this paper, a brief overview of typical DL models is presented first. This is followed by a systematic review of pixel‐wise and scene‐wise RS image classification approaches that are based on the use of DL. A comparative analysis regarding the performances of typical DL‐based RS methods is also provided. Finally, the challenges and potential directions for further research are discussed
AB - Remote sensing (RS) image classification plays an important role in the earth observation technology using RS data, having been widely exploited in both military and civil fields. However, due to the characteristics of RS data such as high dimensionality and relatively small amounts of labeled samples available, performing RS image classification faces great scientific and practical challenges. In recent years, as new deep learning (DL) techniques emerge, approaches to RS image classification with DL have achieved significant breakthroughs, offering novel opportunities for the research and development of RS image classification. In this paper, a brief overview of typical DL models is presented first. This is followed by a systematic review of pixel‐wise and scene‐wise RS image classification approaches that are based on the use of DL. A comparative analysis regarding the performances of typical DL‐based RS methods is also provided. Finally, the challenges and potential directions for further research are discussed
KW - convolutional neural network
KW - deep belief network
KW - deep learning
KW - pixel-wise classification
KW - remote sensing image
KW - scene classification
KW - stacked auto-encoder
UR - http://www.scopus.com/inward/record.url?scp=85047519048&partnerID=8YFLogxK
U2 - 10.1002/widm.1264
DO - 10.1002/widm.1264
M3 - Review article
VL - 8
JO - WIREs Data Mining and Knowledge Discovery
JF - WIREs Data Mining and Knowledge Discovery
IS - 6
M1 - e1264
ER -