Research And Application Of Key Technologies For Remote Sensing Water Quality Inversion Based On Big Data Platform

Remote sensing water quality inversion has great potential in water quality monitoring,but with the rapid growth of inversion related data,especially remote sensing images,standalone data storage and processing methods are facing challenges,while big data storage and parallel processing technology provides new ideas for remote sensing inversion of water quality under massive data scenarios.The thesis explored the design and implementation of remote sensing water quality inversion system based on big data platform.The paper provides basic data support for the model by building a big data platform,studies a variety of regression methods to determine the most suitable inversion model,and uses Spark to realize the parallelization of the inversion model.The thesis finally applied the inversion model to the multi-phase remote sensing image of Futou Lake to explore the temporal and spatial distribution of water quality in Futou Lake.The main contents and results of the research are as follows:(1)Using big data components such as HDFS,Accumulo,Elastic Search,etc.,a basic data platform was built to store and retrieve the raw data related to water quality inversion,remote sensing image tile-pyramid data,water quality data,water body vector boundaries and other heterogeneous data.The platform provides data support and uses Spark to realize the parallel construction of remote sensing image tile-pyramid.(2)The stepwise multiple linear regression,SVR,and GBDT methods were selected to construct inversion models for the measured water quality data of Futou Lake and the quasi-synchronous Sentinel-2 remote sensing images.Taking the coefficient of determination R~2 and the root mean square error RMSE as the evaluation indicators,it is concluded that the GBDT method is most suitable for constructing the Chl-a concentration inversion model of Futou Lake.The R~2 of the constructed model on the test set is 0.6876,and the RMSE is 0.0032mg/L.(3)Using Spark to implement the GBDT method,a parallel remote sensing inversion model of Chl-a concentration was constructed.The inversion model was applied to the multi-phase remote sensing images of Futou Lake,and the temporal and spatial distribution of Chl-a concentration in Futou Lake during the corresponding period was analyzed.In terms of time,the concentration of Chl-a fluctuated and declined from August to December,and showed an upward trend from December to January.The overall trend was summer>autumn>winter.Spatially,the concentration of Chl-a showed an increasing trend from south to north in August when the flow was relatively large,and a decreasing trend from south to north from December to February of the following year.The paper chose Futou Lake as a pilot study to verify the feasibility of water quality inversion based on a big data platform,which can provide a reference for large-scale and long-term water quality monitoring in rivers and lakes.