摘要
三维模型语义标注目的是自动得到最合适描述模型的标注词集合,“语义鸿沟”的存在有必要进一步提高标注效果.传统的聚类算法只是针对模型的内容特征,并没有考虑模型的语义信息.针对以上缺点,本文提出一种基于语义与特征相结合的聚类分析方法,并在此基础上通过相关反馈实现三维模型语义标注,最后结合通用本体库WordNet扩展标注词从而达到较好的语义标注效果.用测试数据集PSB(Princeton Shape Benchmark)验证了该方法取得较好的标注效果.
In order to automatically get a set which is the best description of 3D model annotation, "Semantic gap" makes the necessary to further improve annotation performance. However, traditional cluster algorithms only took into account the content feature without semantic information. Therefore, this paper proposes art annotation method based on cluster analysis and relevance feedback, and finally use WordNet dictionary to expand users annotation terms. The test result on the PSB data set has shown that the method has achieved a better annotation result.
出处
《山东师范大学学报(自然科学版)》
CAS
2015年第4期50-54,共5页
Journal of Shandong Normal University(Natural Science)
关键词
语义标注
聚类分析
相关反馈
semantic annotation
cluster analysis
relevance feedback