The Application Of Hyperspectral Data In Shallow Bottom Classification And Water Depth Inversion

Ocean remote sensing is the focus of current scientific research, hyperspectral remote sensing is a new technology in the field of ocean remote sensing, there are great practical value and academic value in studying on the methods for extracting information from hyperspectral data. This paper research the methods for identification of shallow bottom types and deriving water depth based on spectral derivation technology and artificial neural network.Firstly, this paper expatiate the basic theories of ocean remote sensing, describe the existing methods of remote sensing image classification. And then, basing on spectral characteristics of radiation attenuation characteristics and shallow bottom of light in the water, combining with theoretical model and empirical parameter, realizing the construction of semi-theoretical semi-empirical model. As a basis for generating the three different types of sediment in the shallow reflectivity simulation data for this paper provided a useful data base.Secondly, this paper uses fourth-derivative on processing sediment reflectivity data simulate by semi-theoretical semi-empirical model, standing out the nuances of different sediment spectrum. The results show that after treating the sediment reflectance spectrum with fourth-derivative, the slope values between the two bands can distinguish substrate types. Then basing on data of simulation sediment experiment, proving the practicability and accuracy of fourth-derivative algorithmic.Thirdly, this paper studies the sediment classification and water depth deriving using artificial neural network. Selecting the appropriate number of hide layer and hide layer neurons to simulate the actual measuring data, the results show that the neural network can inverse substrate type and water depth correctly.Finally, this paper uses a water depth inversion arithmetic which base on sediment classification for the situation of mix substrate. Using different water depth inversion model to compute water depth. By comparing the inversion results with the error of measure depth, prove sediment classification based on water depth inversion algorithm is more applicable to mixed substrate conditions. And computing the water depth of shallow sea of some area in our country which base on the algorithm, finding the accuracy of water depth can improve by the sediment classification.