Journal of Geo-information Science

Water Body Extraction from Remote Sensing Images Based on Improved HarDNet-MSEG

GUO Huilin, XIE Yuanli, HU Lifa, YONG Jiale, LI Yunmei, SUN Shaoqi

Journal of Geo-information Science, 2024, 26(7): 1745-1762. DOI: 10.12082/dqxxkx.2024.230656

Fig. 2 Different locations of HAM in the network

Other figure/table from this article

Fig. 1 Hybrid Attention Mechanism
Fig. 3 Overview of the Qinghai-Tibet Plateau
Tab. 1 Row and column numbers of acquired images
Tab. 2 Experimental environment
Fig. 4 Loss, MIoU, FWIoU, and PA curves of different models
Tab. 3 Training results of different models （%）
Fig. 5 Extraction results of different segmentation algorithms
Tab. 4 Training results of different semantic segmentation networks
Tab. 5 Results of water extraction by different methods
Fig. 6 Comparison of extraction results of different methods
Tab. 6 Extraction accuracy of HarDNet-MSEG-HAM1 on other datasets （%）
Tab. 7 Total area of lakes in the inland lake areas during the four time periods
Fig. 7 Changes in the area of lakes in the instream flow basin of the Tibetan Plateau from 2013 to 2022