Calculation of fisher information matrix and its application in methods of recognition and classification for remote sensing of earth

The application of Fisher information matrix in methods of recognition and classification for remote sensing of Earth is considered. The determination and calculation algorithm of Fisher information matrix for the case of independent observed objects of probability learning sample are presented. It is shown the use of the Fisher information matrix in statistical theory for the estimation of precision of a recognition and classification, namely, Fisher information matrix is applied for the calculation of mean square errors of lower confidence bound in Bayesian probability, of classification quality criterion of probability learning sample, and for construction of optimal method of pattern classification on such factors as “set of object characteristics” and “classification rule”. The optimization of the method is carried out on continuous optimization trajectory on the hyperplane of two factors “set of object characteristics” and “classification rule”. The ways of improvement of the presented approach are proposed.