A novel semantic segmentation approach based on U-Net,WU-Net,and U-Net++deep learning for predicting areas sensitive to pluvial flood at tropical area  

作　　者：Laura Melgar-García Francisco Martínez-álvarez Dieu Tien Bui Alicia Troncoso 

机构地区：[1]Data Science&Big Data Lab,Pablo de Olavide University,Seville,Spain [2]GIS Group,Department of Business and IT,University of South-Eastern Norway,Gullbringvegen 36,B?i Telemark,3800,Norway

出　　处：《International Journal of Digital Earth》2023年第1期3661-3679,共19页国际数字地球学报（英文）

基　　金：The authors would like to thank the Spanish Ministry of Science and Innovation for the support under the projects PID2020-117954RB-C21 and TED2021-131311B-C22.

摘　　要：Floods remain one of the most devastating weather-induced disastersworldwide, resulting in numerous fatalities each year and severelyimpacting socio-economic development and the environment.Therefore, the ability to predict flood-prone areas in advance is crucialfor effective risk management. The objective of this research is to assessand compare three convolutional neural networks, U-Net, WU-Net, andU-Net++, for spatial prediction of pluvial flood with a case study at atropical area in the north of Vietnam. They are relative new convolutionalgorithms developed based on U-shaped architectures. For this task, ageospatial database with 796 historical flood locations and 12 floodindicators was prepared. For training the models, the binary crossentropywas employed as the loss function, while the Adaptive momentestimation (ADAM) algorithm was used for the optimization of themodel parameters, whereas, F1-score and classification accuracy (Acc)were used to assess the performance of the models. The results unequivocally highlight the high performance of the three models,achieving an impressive accuracy rate of 96.01%. The flood susceptibility maps derived from this research possess considerable utility for local authorities, providing valuable insights and informationto enhance decision-making processes and facilitate the implementation of effective risk management strategies.

关 键 词：Flash-flood assessment convolutional networks climate change GIS 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]