Attribute profiles on derived features for urban land cover classification

International audienceThis research deals with the automatic generation of 2D landcover maps of urban areas using very high resolution multispectralaerial imagery. The appropriate selection of classifierand attributes is important to achieve high thematic accuracies.In this paper, new attributes are generated to increase thediscriminative power of auxiliary information provided by remotesensing images. The generated attributes are derived fromthe vegetation index and elevation information using morphologicalattribute profiles. The extended experimental evaluationand comparison of attribute profile-based mapping solutionsis conducted to derive the optimal combinations of attributesrequired for classification and to understand the genericity ofattributes on a range of classifiers, i.e., various combinationsof attributes and classifiers. Experimental results with two highresolution images show that the proposed attributes derived onauxiliary information outperform the existing attribute profilescomputed on original image and its principal components