Application of U-net model to water extraction with high resolution remote sensing data
Ning WANG1,Jiahua CHENG2(),Hanye ZHANG2,Hongjie CAO1,3,Jun LIU3
1. Beijing Unistrong Science and Technology Co., Ltd., Beijing 100015, China 2. East China Sea Fisheries Research Institute, Chinese Academy of Fishery Sciences, Shanghai 200090, China 3. BeiDou Navigation and LBS (Beijing) Co., Ltd., Beijing 100191, China
In this paper, the authors used a U-net model to conduct water extraction, and the result was compared with that of the random forest model. The accuracy of the U-net model was validated by using GF-1 images in Chaohu Lake Basin. The results show that both models are of high accuracy for large area of water body, but random forest model has more spots for small area of water body, and the result of U-net model is more consistent with the manual visual interpretation result. Moreover, the U-net model can effectively remove the shadows of mountains and buildings. The result indicates that U-net model performs better than random forest model with the overall accuracy of 98.69%, Kappa coefficient of 0.95, omission error of 1.90% and commission error of 1.18%. In contrast, the overall accuracy, Kappa coefficient, omission error and commission error of random forest model are about 98.05%, 0.92, 1.61% and 2.99%, respectively. In addition, the classification features for traditional machine learning model are always calculated by manual extraction. However, the inputs for U-net model are the 4 band spectrum data of GF-1 images. These data suggest that the U-net model avoids the process of manually extracting classification features and has a higher degree of automation. It should be noted that the U-net model uses more train samples with less time-consuming. It is believed that this model can significantly improve the surface water detection accuracy and can be used for the automatic renewal of a larger range of water bodies.
王宁,程家骅,张寒野,曹红杰,刘军. U-net模型在高分辨率遥感影像水体提取中的应用[J]. 国土资源遥感, 2020, 32(1): 35-42.
Ning WANG,Jiahua CHENG,Hanye ZHANG,Hongjie CAO,Jun LIU. Application of U-net model to water extraction with high resolution remote sensing data. Remote Sensing for Land & Resources, 2020, 32(1): 35-42.
Wang J J, Ding J L, Zhang C , et al. Method of water information extraction by improve SWI based on GF-1 satellite image[J]. Remote Sensing for Land and Resources, 2017,29(1):29-35.doi: 10.6046/gtzyyg.2017.01.05.
Bi H Y, Wang S Y, Zeng J Y , et al. Comparison and analysis of several common water extraction methods based on TM image[J]. Remote Sensing Information, 2012,27(5):77-82
Mcfeeters S K . The use of the normalized difference water index (NDWI) in the delineation of open water features[J]. International Journal of Remote Sensing, 1996,17(7):1425-1432.
Xu H Q . Modification of normalized difference water index (NDWI) to enhance open water features in remotely sensed imagery[J]. International Journal of Remote Sensing, 2006,27(14):3025-3033.
Ogilvie A, Belaud G, Massuel S , et al. Surface water monitoring in small water bodies:Potential and limits of multi-sensor Landsat time series[J]. Hydrology and Earth System Sciences, 2018,22(8):1-35.
Fisher A, Danaher T . A water index for SPOT5 HRG satellite imagery,New South Wales,Australia,determined by linear discriminant analysis[J]. Remote Sensing, 2013,5(11):5907-5925.
Feyisa G L, Meilby H, Fensholt R , et al. Automated water extraction index:A new technique for surface water mapping using Landsat imagery[J]. Remote Sensing of Environment, 2014,140(1):23-35.
Wang S D, Baig M H A, Zhang L F , et al. A simple enhanced water index (EWI) for percent surface water estimation using Landsat data[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015,8(1):90-97.
Guo Q D, Pu R L, Li J L , et al. A weighted normalized difference water index for water extraction using Landsat imagery[J]. International Journal of Remote Sensing, 2017,38(19):5430-5445.
Diao S J, Liu C L, Zhang T , et al. Extraction of remote sensing information for lake salinity level based on SVM:A case from Badain Jaran desert in Inner Mongolia[J]. Remote Sensing for Land and Resources, 2016,28(4):114-118.doi: 10.6046/gtzyyg.2016.04.18.
Feng Q L, Liu J T, Gong J H . Urban flood mapping based on unmanned aerial vehicle remote sensing and random forest classifier:A case of Yuyao,China[J]. Water, 2015,7:1437-1455.
Badrinarayanan V, Kendall A, Cipolla R . SegNet:A deep convolutional encoder-decoder architecture for image segmentation[EB/OL]. (2015-11-02). https://arxiv.org/abs/1511.00561.
Noh H, Hong S, Han B . Learning deconvolution network for semantic segmentation[C]// IEEE International Conference on Computer Vision.Santiago:IEEE, 2015: 1520-1528.
Chen L C, Papandreou G, Kokkinos I , et al. DeepLab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016,40(4):834-848.
Ronneberger O, Fischer P, Brox T . U-Net:Convolutional networks for biomedical image segmentation[C]// International Conference on Medical Image Computing and Computer-Assisted Intervention.Cham:Springer, 2015: 234-241.
伍广明, 陈奇, Shibasaki R , 等. 基于U型卷积神经网络的航空影像建筑物检测[J]. 测绘学报, 2018,47(6):864-872.
Wu G M, Chen Q, Shibasaki R , et al. High precision building detection from aerial imagery using a U-Net like convolutional architecture[J]. Acta Geodaetica et Cartographica Sinica, 2018,47(6):864-872.
Chen C, Zhou Y J, Li Y X , et al. Remote sensing image registration based on U-net segmentation and the HEIV model[EB/OL]. (2018-11-30). https://kns.cnki.net/KCMS/detail/31.1289.tp.20181129.1323.005.html.