In this article, a segmentation and texture-based image classification method is proposed to extract the urban areas of the municipalities of Sintra and Cascais using a Landstat enhanced thematic mapper plus (ETM+) image. This method is compared with other classification methods, among some of the most used in remote sensing: normalized difference built-up index (NDBI), maximum likelihood, clusters, and neural networks. The results obtained by the proposed method, using overall accuracy and kappa indices, are superior to the ones produced by the other methods. The results further indicate that the urban class, of heterogeneous spectral composition, is difficult to classify by purely spectral methods that do not consider the structure of the image.
|Translated title of the contribution||Delineation of urban areas from a Landsat ETM+ image: comparison of classification methods|
|Number of pages||9|
|Journal||Canadian Journal of Remote Sensing|
|Publication status||Published - 2007|