**6. Conclusions**

In this study, we proposed a new end-to-end cultivated land extraction algorithm, the highresolution U-Net (HRU-Net). Compared with the original U-Net, the HRU-Net had two improvements: (1) it improved the skip connection structure, and (2) it used the idea of deep supervision to modify the loss function. We tested the proposed method and compared it with the U-Net++, U-Net, and the RF on three Landsat TM datasets with different spectral band combinations and drew the following conclusions:


The HRU-Net model presented in this study demonstrated good performance in extracting the target with high edge details and high intra-class spectral variation. This model can be further used to extract the target within these characteristics. The model introduced in this study can be extended or combined to more other high spatial resolution satellite data, such as Sentinel-2, GF1, and GF2.

**Author Contributions:** Data curation, W.X. and X.W.; formal analysis, W.X.; investigation, L.S.; methodology, W.X. and S.G.; project administration, J.C.; software, X.Z., Y.X. and Y.S.; supervision, X.D., S.G. and L.S.; writing—original draft, W.X.; writing—review & editing, S.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the National Key Research and Development Program of China (Project No. 2017YFB0504203), the Natural science foundation of China project (41601212, 41801358, 41801360, 41771403), and the Fundamental Research Foundation of Shenzhen Technology and Innovation Council (JCYJ20170818155853672).

**Acknowledgments:** The authors thank C. Ling from SIAT for discussion and suggestions.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
