Ontology-based object recognition of car sides

Object recognition is one of the most classic problem in computer vision where several techniques in classification have been proposed. In this work, we investigate a new promising improvement on a classification problem using multiple feature types. We assume that different features are suitable for different objects and the occurrence of particular objects usually relate to each other via particular relations. One possibility of finding such a special connection is to employ the advance of semantic technologies. Here, we investigate a new detection framework capable of detecting composite object instances. Unlike most of the previous works which attempts to detect an individual object as a whole, we aim at identifying its composition using an OWL ontology. To this end, low-level features obtained from an image are converted to ontological assertions, and the semantic reasoning of realization is employed to check the type of objects present in the image.