A Shape Ontology Framework for Bird Classification

Current research on shape based classification has been generally aimed at utilising various visual features. Previous research has shown that the existing knowledge in a specific domain can assist in understanding the image content. Ontologies are currently being used for explicit representation of the domain knowledge. In this paper, two contributions are presented: 1) a shape ontology framework which constitutes both domain and shape ontologies and in which domain and shape ontologies are mapped to each other. 2) A new approach for automatic construction of shape ontology. The experimental results are promising. Future work will focus on validating the framework and automatic method of the shape ontology construction for a much larger dataset.