Department of Pattern Recognition

(eng): A big number of monitored features in examined objects often complicates a technical realization of decision-making and extends the time necessary for providing a decision. It is possible to decrease dimensionality of the tasks and along with not to decrease a quality of decision-making. The main subject of this contribution relates to some of the possible approaches. Basis of these methods stays in finding a linear transformation of original m-dimensional space of features into a new n-dimensional feature space where n pound m. New features are arisen by suitable linear combination of original features and they are descending sorted according to their variance. The contribution describes experience and results obtained with the decreasing of feature space and the classification of sets of the objects which where represented by real pictures of the Earth surface.

(cze): Velké množství příznaků komplikuje realizaci rozpoznávaciho procesu a prodlužuje čas rozhodovacího procesu.