Fitting Estimation Of Variance And Covariance Components Under Parametric Cross Correlation

In the process of measurement data,In the regular collocation of variance and covariance,people always assume that stochastic signals have isotropic random process,and does not consider the correlation between the random signal and the observed noise in the process of calculation,but in reality the random signal anisotropy is universal,and there are also correlative relationship between stochastic signals and observation noise.If the correlation is not considered in the process of fitting the estimation,the result of the operation is not optimal.Therefore,the theory and methods of various collection appeared one after another.In order to obtain the best estimate,this thesis aims to use the least-squares collocation as research model to calculate the parameter X and valuation Y in the premise of considering the relationship between the observed noise and the random signal,as well as between the measured signal and unmeasured signal.In addition,this thesis also adopts the variance-covariance of the observed noise and the random signal to ensure the random model.After getting the analysis formulas,this thesis adopts the minimum norm quadratic unbiased estimator to coordinate the weight ratio between the observed noise and the random signal in the fitting estimation model.And afterwards,using MATLAB soft to the fitting estimation.Aiming to compare the results of observation under the independent situation and parametric situation,analyzes the difference,and explains the practicability and superiority of the algorithm.