Classification Of Hyperspectral Images Without Calibration Information

Hyperspectral image contains abundant spatial image information and spectral radiation information.It is widely used in environmental monitoring,civilian applications,and natural resource exploration.Currently,the ablity of analyzing and processing remote sensing data is still limited.How to extract useful information from remote sensing data become a primary problem of quantitative remote sensing.This thesis mainly discuss the inversion algorithm of hyperspectral reflectance.The structure of this thesis is as follows:(1)The thesis introduces the research background and significance of hyperspectral remote sensing image.In addition,current research status of radiometric calibration and atmospheric correction is outlined.(2)In this thesis,we describe the inversion algorithm of traditional reflectance.The principle of two kinds of correction algorithms based on physical radiative transfer model and regression statistical model are introduced.Furthermore,The advantages and disadvantages of different algorithms are compared.(3)Due to the lack of calibration information and atmospheric optical parameters,no systematic and effective methods have been specifically developed for hyperspectral targets recognition,which has restricted its application in quantitative remote sensing.A method using particle swarm optimization to choose the parameters in 6S model is proposed,and is applied to hyperspectral target recognition.It solves the multi-parameter nonlinear optimization problem,and the effectiveness of the method is verified by the hyperspectral data provided by the Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Sciences.(4)In the absence of synchronized observation of optical parameters,the non-linear regression model method has not been widely applied.Therefore,it is an important research topic for the statistical model method through the study of radiative transfer process and the measured scatter plot of data distribution.In this thesis,a set of hyperspectral remoting image has been used to test the non-linear regression method for atmospheric correction.Simulations show that: the logarithmic model is superior to the traditional empirical linear models.