Land cover data from Landsat single-date imagery: an approach integrating pixel-based and object-based classifiers

This paper proposes an approach for generating land cover information from single-date Landsat images integrating pixel-based and object-based classifiers in two study areas: the province of Oristano and the region of Campania. The process consisted in: a) pre-processing; b) segmentation; c) classification based on radiometric properties and integration with textural properties and vegetation indices. A good overall classification accuracy was obtained at the first (Oristano: 87%, Campania: 88%) and at the second (Oristano: 74%, Campania: 82%) CORINE Land Cover level. Results highlight the potential of the method to be replicated in time and space in the perspective of a semi-automatic and cost-efficient transferability of the procedure.