The use of fuzzy set theory in remote sensing pattern recognition

The ultimate goal of urban land cover mapping using high-spatial-resolution images acquired by earth-orbiting sensors is to monitor both the extent of urban areas and their composition in terms of land use and, thus, to obtain accurate and consistent information on urban areas. Conventional classification techniques, however, have not always provided sufficiently accurate thematic maps. A fundamental problem is that conventional pattern classification techniques used are hard (crisp) techniques in which each image pixel is associated with only one class throughout the classification. Although this may be reasonable with some relatively fine spatial resolution remotely sensed data sets, coarse spatial resolution satellite sensor imagery such as that acquired by the Thematic Mapper are usually dominated by pixels of mixed land cover composition in urban areas. Failure to accomodate for mixed pixels will result in a poor representation of land cover distribution and incorrect estimates of class extent derived from it. This contribution shows how fuzzy set theory may be incorporated into the classification process, namely at the class membership level and at the output level with a measure of fuzziness to make aware of the vagueness of land cover classification obtained.