NON-PARAMETRIC APPROACH FOR PRELIMINARY PROCESSING OF EARTH REMOTE SENSING DATA

Abstract

Researches described in the paper are aimed at studying the methods of data preprocessing from a sample of observations of a system characterized by input-output values of variables. We consider the data containing omissions and outliers. Algorithms for leveling outliers in a sample of observations, as well as algorithms for filling data gaps are presented. In addition, it is implemented a data repair algorithm that is able to recover lost values (outliers) after their exclusion. Our studies are useful in geographic information systems or in the analysis of information received from satellites during remote sensing of the earth.