Urban Area Classification By Multispectral SPOT Images

Two different classification methods on SPOT MS images were investigated with the goal of fast updating of thematic maps. The study focused on recent urban development of the medium-sized 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 cooccurrence matrix on SPOT band 3 images. Results show that the classification method using the region-growing technique gave satisfactory results in terms of both classification precision and computation time. The method based on textural analysis, on the contrary, is of no practical applicability for SPOT scene classification. (I.E.)