Land cover class extraction in GEOBIA using environmental spatial temporal ontology

Very high spatial resolution (VHSR) remote sensing imaging brings challenges and opportunities to intelligent autonomous interpretation of spatial data due to detailed information available in such images. Accurate extraction of information relies on expert knowledge which can be represented by an Ontology. Within the Geographic Object-Based Image Analysis (GEOBIA) framework, Ontological implementation is recently been started which has created different avenues of novel applications. In this paper, we have developed an Environmental Spatio-temporal Ontology (ESTO), using five different publicly available environmental data sources namely SILO, AWAP, ASRIS, CosmOz, and MODIS, where knowledge was integrated and captured in multiple-scales using resource description framework (RDF). RDF representation made the ESTO very effective way to publish on Linked Open Data Cloud (LOD) environment. ESTO and the RDF adaptation helped for the human-computer interaction with spatial data whereas an automated approach for object interpretation has also been developed. Our Ontological approach integrates thematic with the spatial semantics for the GEOBIA framework. This study tested a WorldView-2 imagery of Hobart, Tasmania, Australia in depicting land cover classes and effectiveness of ESTO for GEOBIA framework.