URBAN AREA CLASSIFICATION BY MULTISPECTRAL SPOT IMAGES

Two different classification methods on SPOT multispectral (MS) images were investigated with a goal of fast updating of thematic maps. The study focused on recent urban development of the medium-size town of Modena in northern Italy for which a thematic map produced in 1984 by aerophotogrammetric detection was available. Of the two classification schemes tested, the first, based on MS response in all three bands, makes use of a modified region-growing technique originally proposed for region extraction from aerial photographs, while the latter measures textural characteristics by means of the co-occurrence matrix on SPOT band 3 images. Results show that the classification method using region-growing technique gave very satisfactory results both in terms of classification precision and computation time; the method based on textural analysis, on the contrary, is of no practical applicability for SPOT scenes classification. © 1990 IEEE