A Study Of Remotely-sensed Data Processing In Bathymetry

Shallow water bathymetry is the hot spot and difficulty in the field of marine surveying and mapping.Limited by the depth of the draught,the traditional shipborne sonar system is inefficient and difficult to implement in the shallow sea area.Optical remote sensing methods have the advantages of low cost and high efficiency,and they are most important supplementary means of shallow bathymetry.Optical remote sensing mainly includes three aspects,water depth inversion from optical remote sensing image,two media photogrammetry and airborne lidar bathymetry.In this paper,the data processing method of optical remote sensing bathymetry is studied.It mainly includes water depth inversion from satellite multispectral imagery,satellite stereopair two media photogrammetry,and airborne lidar bathymetry.The main content and innovation points of this paper are as follows:1.Water depth inversion from satellite multispectral imagery.The water depth inversion model can be established by using the depth sample points and the optical remote sensing image.Firstly,the principle and empirical models of multispectral imagery based depth inversion are introduced.Then,the data processing flow of multispectral depth inversion is analyzed.The paper focuses on the pretreatment of marine optical remote sensing image,and it studies the influence of sample point number on bathymetry accuracy.Several empirical depth inversion models are compared under different water conditions.This article also use the collected different kinds of remote sensing data at home and abroad to perform water depth retrieval experiment,to compare the bathymetry ability of different satellite imagery and to study the influence of image indicators to depth retrieval accuracy.In the end,the traditional global inversion model is improved to the local adaptive inversion model,and the inversion precision of the dual-band ratio model is improved to a certain extent.2.Satellite two media photogrammetry.This paper studies how to extract geometric information from satellite stereopair to calculate the shallow water depth.Firstly,the basic principle of two media photogrammetry based on ray refraction is introduced.Then,the refraction correction model which corrects the straight line intersection point to the broken line intersection point is derived.In addition,the error source of satellite two media photogrammetry is analyzed systematically.In this paper,a complete two media photogrammetry workflow is presented,including radiometric preprocessing,geometric model construction,water and land separation,surface elevation calculation,refraction correction,etc.Finally,we use worldview-2 multispectral stereo images to verify the bathymetrical algorithms proposed in this paper.Experimental results show that the bathymetry accuracy of the satellite two media photogrammetry technology is at the same level and even higher than that of multispectral depth inversion method.Targeted at two-media photogrammetry technology,this paper achieves innovations at the processing workflow and the algorithm level.For example,an epipolar image generation algorithm based on the linear characteristics is proposed,and the sparse matching and intensive matching are combined to obtain reliable underwater parallax.3.Airborne lidar bathymetry.This paper studies how to construct the laser sounding geometric model and how to extract water surface and underwater information from the waveform laser echo data.The basic principle of dual-frequency laser sounding and equipment used in this paper are also first discussed.Then,the lidar data processing workflow is divided into three phases: geometric model construction,point cloud's water land separation and echo wave recognition.In the geometric model,the geographic coordinate transformation formula of laser points is derived.In the aspect of water land classification,we predict the class labels based on support vector machine using the feature selected before.For echo wave recognition,the nonlinear least squares solution of the mixed Gaussian model is derived in detail.The core problem of the lidar bathymetry is to deal with the shallow water area of the mixed waveform and the deep water area of the weak underwater signal.The paper proposes a method to simulate the surface component of green shallow water by using the surface near infrared channel,to separate the surface and bottom signal in the shallow green laser echo.Multi-channel echo calculation for deep water area is used to improve the success rate of water depth measurement.At last,the lidar bathymetry algorithms are tested by single beam sonar data,and results show that the RMSE error is better than 0.2m,which fully meets the precision requirement of the water depth measurement in the offshore development activities.