Research And Application Of Remote Sensing Water Quality Inversion Algorithm For Inland Waters

With the great development of remote sensing technology,it is possible to achieve low-cost,large-scale,high temporal resolution and contactless water quality inversion by using remote sensing technology.Aiming at the problems of high data acquisition cost,poor model generality,and inadequate combination of theoretical research and practical application in the remote sensing water quality inversion of inland waters,this thesis studies the research on algorithms of remote sensing water quality inversion for inland waters.The research contents of this thesis are summarized as follows:(1)Aiming at the problem of low applicability of quasi-analytical algorithm(QAA)in water quality inversion for inland waters,a semi-analytical method for inland waters based on the modified QAA is proposed.Based on the homogeneity and rationality of the inherent optical properties between adjacent points in a specific water area,the modified QAA realizes unsupervised modification of the QAA to adapt to the current area.Then,absorption coefficients are retrieved by the modified QAA and inversion model of suspended matter concentration is constructed by using selected band subset.Furthermore,the spatial distribution map of total suspended matter(TSM)is obtained by using normalized difference total suspended matter concentration index(NDTI)model with the best inversion effect.(2)Aiming at the problem of dimensionality curse caused by the strong correlation and high redundancy of hyperspectral data,a band selection method based on pre-trained neural network(NNBS)is proposed for water quality inversion of inland waters.Due to the fact that parameters in a neural network can express the importance of features,band selection for hyperspectral data is performed by training neural network.Then,the random forest and neural network are employed to build inversion models of TSM and chlorophyll concentration by taking the selected band set as features of input data.The experimental results validate that the proposed NNBS based on the L1 norm and the L2 norm can obtain a more informative band subset,and achieve better results when using neural networks as a downstream model to construct TSM and chlorophyll concentration inversion models.(3)Combining the aforementioned theories and methods,water management and monitoring software based on remote sensing water quality inversion is designed under the browser/server(B/S)framework.The software comprises several functional modules such as system management,data management of waters,water quality inversion,water quality data analysis,and management of water affairs.