An ontology-driven approach for the extraction and description of geographic objects contained in raster spatial data

► We propose a ontology-driven methodology for feature extraction from raster data. ► Ontologies are created using the GEONTO-MET methodology. ► We present a novel semantic decomposition algorithm based on an isotropic space. ► The semantic decomposition algorithm finds clusters of pixels sharing common features. ► Clusters or regions get semantic descriptions based on the ontologies. In this paper, we present FERD, a methodology aimed to automatically identify, extract and describe relevant spatial objects contained in raster spatial datasets. Our objective is to provide a set of computational tools capable of finding landforms contained in the datasets that match human-friendly descriptions such as “In this model there is a mountain having a maximum altitude of 302m, located between coordinates (19.09383°N, 99.85541°W) and (19.09393°N, 99.85554°W)”. The proposed methodology is composed of three main stages: in the first stage (conceptualization), the knowledge domain is represented by means of ontologies. In the second stage (synthesis) a novel semantic decomposition algorithm is used to identify and extract relevant spatial objects from the spatial dataset. In the last stage (description), the geographic objects extracted in the second stage are mapped to concepts (objects of the knowledge domain) generated in the first stage. The final result is a set of metadata that describes the geomorphologic objects contained in the raster dataset.