Dependence of the results of classification of multispectral images of forest vegetation on wavelet-transform parameters

Abstract

The problem of classification of space multispectral images of forest vegetation is considered. Features based on the wavelet transform and a classification method involving consideration of the importance of each feature are investigated. Dependence of the classification results on the wavelet function, the level of the transform, and the parameter of the classification method — the number of segments of the range of features — is given. Results of classification of multispectral images of six classes of forest vegetation in images obtained by a Rapid Eye shooting system are presented.