Researches on Semantic Annotation and Retrieval of 3D Models Based on User Feedback
As an important part of multimedia retrieval, researches on 3D model retrieval concentrate on the shape-based retrieval method. It is a promising way to improve retrieval performance by adopting semantic information. At present, semantics of an object is usually represented by several keywords. However, acquiring each 3D model’s semantics is very difficult and expensive. To solve the problem, the paper proposes to describe a 3D model’s semantics based on its relationship with the semantics of the others, and states an automatic semantic annotation based on noisy user feedbacks. The paper first analyzes the semantic relationship reflected by user feedbacks. Then, the semantic relationship is treated as one 3D model’s semantic property and is adopted in clustering to detect semantic groups that is named as semantic community. Thirdly, based on the semantic community, the semantics for models is automatically and efficiently annotated based on semantic keywords of a few 3D models. Finally, a retrieval mechanism with long-term semantic learning ability is proposed. The experiments performed on Princeton Shape Benchmark show that the proposed method achieves good performance not only in semantic clustering and annotation but also in semantic retrieval.
Tianyang Lu Shaobin Huang Peng Wu Yeran Jia
College of Computer Science and Technology, Harbin Engineering University Harbin, China
国际会议
Sixth International Conference on Semantics,Knowledge and Grids(第六届语义、知识与网格国际会议 SKG 2010)
宁波
英文
211-218
2010-11-01（万方平台首次上网日期，不代表论文的发表时间）