Textural Classification of Forest Types from Landsat 7 Imagery

Abstract

A method was developed for using textural, brightness, and spectral indices for the classification of forest valuation tracts on the basis of space imagery. The textural characteristics were determined by applying a discrete wavelet transform and with use of a textural matrix, on the basis of which the following textural indices were determined: energy, standard deviation, inertia, entropy, homogeneity factor, cluster shade (asymmetry), cluster prominence, and cluster correlation information measure. An unsupervised classification of valuation tracts was made for a space image of a test area within the Bulun forested sector of the Zhigansk forest management area registered by the Landsat 7 satellite. The use of textural indices alone for classification purposes proved to be inadequate. Satisfactory results were obtained only with the joint use of both textural and brightness spectral indices.