원문정보
초록
영어
Despite the vast research amount on the analysis and retrieval sematic images, there are still significant challenges worthy of address. This paper proposes an ontology framework to analyze and retrieve text and image based semantic search. This framework can be described in three main processes. Firstly, the Query Engine process constructs the input image query in SPARQL language. Secondly, the process of matching module is to retrieve the most affined images based on the compliance with input query. This process extracts the shape features of image's objects via chord-length features. Furthermore, the ontology manger process inserts the new relevant object's features in ontology knowledge base. Finally, the ranking module process is to classify the images which displayed in descending ordered based on matching values. Our experiment on a trained benchmark mammals shows that the proposed framework is more vigorous and yields favorable results when applying auspicious with large number of tested images without sacrificing real-time performance.
목차
1. Introduction
2. Related Literature
2.1. Chord-length Function
2.2. Conditional Random Fields Classifier
3. Suggested Approach
3.1. Query Engine (QE)
3.2 Matching Module (MM)
3.3 Ontology Manager (OM)
3.4. Ranking Module (RM)
4. Experimental Results
4. Summary and Conclusion
References