New trend in object oriented image analysis - ontology

Abstract

Spatial data is used for many purposes in city administration such as solving problems of living areas, establishing healthy and sustainable cities and establishing the infrastructure of smart cities. For this reason, collection, processing, evaluation and transformation of spatial data are important for quick and accurate decisions of city administrators. In recent decades, different methods and algorithms have been developed in order to improve and optimize object extraction techniques for spatial data. However, spatial data have different technical characteristics (geometric, radiometric, temporal resolution, etc.) since they are mostly obtained from different data sources. For that reason, spatial data shows a heterogeneous structure in terms of spatial semantics. This heterogeneity creates problems in the conceptualization of expert knowledge, interoperability and reusability. Ontology has become a current research topic in the elimination of problems related to heterogeneity in object extraction studies for helping to create a conceptualized expert knowledge, semantically presented and linked with each other. In this study, ontology driven object extraction is aimed by using LiDAR system data of Evrencik/ Kirklareli/Turkey. For this purpose, object classification was made by object oriented image analysis method and fuzzy logic classification. In this study, accuracy analysis was performed and the overall accuracy value 0.88 and the kappa value 0.74 obtained. In addition, ontology was developed by defining conceptual class definitions, object and data properties, rules and axioms. Semantic infrastructure model was established by integrating classified image objects and developed ontology.