INFORMATION SYSTEM FOR NATURAL OBJECTS STATE ESTIMATION ACCORDING TO REMOTE SENSING DATA BASED ON NONPARAMETRIC DECISION-MAKING ALGORITHMS
The unconventional information system for the estimation of natural objects condition according to the remote sensing data is considered. It is based on non-parametric algorithms of pattern recognition with gradations of advantage. Proposed is an optimization methodology for nonparametric classifiers based on the fuzziness ratios of kernel functions. The functionality of the developed software tools allows solving the problems of spectral data primary processing, testing hypotheses about the distribution of multidi-mensional random variables, estimating natural objects condition and their spatial distribution. The software tools are implemented in Delphi environment. The initial information of remote sensing data is presented in the form of Microsoft Office Excel spreadsheets. The proposed software tools are used to estimate the forest stands condition according to remote sensing data. The efficiency of the developed software and the software product ERDAS Imagine is compared.
Keywords: nonparametric decision-making models; pattern recognition; estimation of research objects condition; hypothesis testing; multidimensional random variables; kernel density estimation; selecting bandwidth; spectral data; remote sensing; forest areas