Research On Bathymetry Algorithm Without Ground Data Based On Multispectral Remote Sensing Image For Shallow Water Of Islands

As an important part of hydrological data,water depth can provide data support for human activities such as ship navigation and underwater terrain monitoring.Furthermore,flourishing of the aerial remote sensing technology has made large-scale,high temporal resolution and contactless bathymetric inversion in neritic areas possible by means of remote sensing.A large number of scholars have searched on remote sensing bathymetric inversion in neritic areas,but only a few studies focus on bathymetric inversion w/o ground data.As a result,the paper focuses on a neritic bathymetric inversion w/o ground data method of multi-spectral remote sensing in Ocean Island Reefs,with exploration from following aspects:(1)An review of remote sensing bathymetric inversion algorithm.A variety of bathymetric inversion algorithms with ground under multispectral conditions was reviewed in the paper,including single-band,dual-band and logarithmic ratio bathymetric inversion algorithms.Based on an analysis of the bathymetric inversion algorithm w/o ground data under hyperspectral conditions,that is,the semi-analytical model,their advantages and disadvantages as well as applicable scenarios were analyzed by comparing the above algorithms.(2)Research on the bathymetric inversion method w/o ground data based on machine learning.In terms of issued related to neritic bathymetric inversion w/o ground data in Ganquan Island reefs of the South China Sea,various machine learning models were trained by using remote sensing images and bathymetric data from other waters,such as linear regression,support vector regression,integrated learning and neural networks,to achieve bathymetric inversion w/o ground data in Ganquan Island.Experiments demonstrated that different machine learning models work in bathymetric inversion w/o ground data to a certain extent,but the inversion has low accuracy and stability.(3)Based on an in-depth analysis of advantages and disadvantages of the logarithmic ratio model and the semi-analytical model as well as their application limitations,a novel method(L-S model)applicable to four-band remote sensing images to achieve bathymetric inversion w/o ground data was proposed.Optimal inversion parameters were determined through contrastively analyzing bathymetric inversion results of semi-analytical models of randomly selected pixels,and their spectral characteristics,and results were obtained through a logarithmic ratio model based on the optimal parameters.The experiment results of Hainan's Ganquan Island's bathymetric inversion w/o ground data,which were based on a variety of remote sensing data,indicated that in the absence of ground data,the model cannot only achieve similar precision to the bathymetric inversion with ground data,but also has higher stability.(4)Software design and development of bathymetric remote sensing inversion w/o ground data for Ocean Island Reefs have been completed based on the LS model,with detailed design of the software modules,such as image display,sun glint correction,atmospheric correction,bathymetric inversion w/o ground data,and output of darkcolor sequence diagrams of water depth.The application results indicated that the software has an outstanding ability of bathymetric inversion w/o ground data in neritic area of Ocean Island Reefs.