Research On Hydrological Information Extraction And Water Supply In "One Belt And One Road" Region Based On Multi-source Remote Sensing Data

Water is an important resource to ensure human survival and economic and social development.Research on water supply by remote sensing and GIS technology has always been a hot issue in the field of hydrology,especially for the“One Belt And One Road" region,where water resources are lacking and border conflicts occur from time to time.The research on water supply in "One Belt And One Road" region is an important subject which affects the economic and social development and guarantees the border security of China.At present,the identification and detection of hydrological elements mostly rely on visual interpretation and manual labeling methods,which have a low degree of automation and are difficult to meet the needs of rapid water search in a large range.Therefore,it is urgent to find out a water supply analysis methods with a higher degree of automation.In response to the above problems,combined with deep learning and data analysis theories in remote sensing and GIS,we completed automatic extraction of important hydrological information,ice and snow and springs,and carried out research on water supply guarantee in "One Belt And One Road" region based on multi-source remote sensing data.A prototype system for extracting typical hydrological information from remote sensing is designed.The main contributions of the thesis are as follows:(1)Ice and snow identification from Landsat8 images based on Deep U-Net was designed.Ice and snow is an important surface freshwater resources in the study area,and also an important source of groundwater in the study area.In this paper,the typical hydrological information,ice and snow was selected,and Landsat8 OLI image was used as the data source to establish the ice and snow sample set.Deep learning semantic segmentation network Deep U-Net was applied to complete the automatic identification of ice and snow.The results show that the method can improve the lower the false alarm and achieve high identification accuracy.(2)The spring point detection method based on YOLOV3 using high resolution remote sensing image was designed.Water supply from springs is an important method of groundwater utilization in border areas.Traditional spring detection mainly relies on manual visual interpretation of images and field verification,which is inefficient.This paper established a sample set of springs in the study area based on high-resolution remote sensing images.The deep learning network YOLOv3 was applied to realize the automatic detection of spring point.(3)In view of the problem that the weight determination of evaluation indexes in the current groundwater assessment model was too much affected by human.We analyzed the various factors related to the groundwater,constructed a grid-based evaluation index system for groundwater potential,and an evaluation model for groundwater potential based on the evaluation index system,gridding approach,entropy weight method combined with artificial experience,Gamma transformation,and natural breakpoint classification method was proposed.This evaluation method combined objective weights and subjective experience to reduce the dependence of artificial experience.(4)Furthermore,according to the GIS vector data,considering the terrain conditions and traffic conditions of the study area,the water supply guarantee mode of surface water such as ice and snow,rivers,lakes,and reservoirs is analyzed;Based on the grid evaluation results of the groundwater-rich potential area in the study area,and the principle of delineating groundwater potential area,combined with topography and traffic vector data,a specific analysis of the delineated groundwater potential area is carried out.(5)In response to the needs of large-scale hydrogeological research,we integrated the automated ice and snow range extraction method,automated spring point detection method and grid evaluation method of groundwater rich potential area designed in this thesis,and designed a prototype system for extracting typical remote sensing hydrological information.The system included functions such as data management and display,hydrological information extraction and data processing.The ease of operation of the system reduces the difficulty of applying the remote sensing hydrological information extraction method in practical,and provides a reference for a large-scale water supply research.