Improving the Computational Performance of Ontology-Based Classification Using Graph Databases

Abstract

The increasing availability of very high-resolution remote sensing imagery (i.e., from satellites, airborne laser scanning, or aerial photography) represents both a blessing and a curse for researchers. The manual classification of these images, or other similar geo-sensor data, is time-consuming and leads to subjective and non-deterministic results. Due to this fact, (semi-) automated classification approaches are in high demand in affected research areas. Ontologies provide a proper way of automated classification for various kinds of sensor data, including remotely sensed data. However, the processing of data entities—so-called individuals—is one of the most cost-intensive computational operations within ontology reasoning. Therefore, an approach based on graph databases is proposed to overcome the issue of a high time consumption regarding the classification task. The introduced approach shifts the classification task from the classical Protégé environment and its common reasoners to the proposed graph-based approaches. For the validation, the authors tested the approach on a simulation scenario based on a real-world example. The results demonstrate a quite promising improvement of classification speed—up to 80,000 times faster than the Protégé-based approach.