Abstract: The Princeton Shape Benchmark（PSB） has become one of the most prevailing benchmarks in 3D Model Retrieval community.However，categories similarity is ignored in PSB and thus raises inconsistency between the objective retrieval performance evaluation and the subjective human perception.To address this issue，a new classification method considering categories similarity is proposed in this paper.Based on the proposed classification approach，a novel retrieval performance evaluation measure using similarity sequence analysis is then designed and discussed in details.

Key words: 3d model retrieval, benchmark, similarity sequence analysis

摘要： 普林斯顿形状基准数据库已经成为当前三维模型检索研究人员使用最为广泛的基准数据库之一。但是该基准数据库存在着一个重要的缺陷，那便是模型类别之间的相似性没有被考虑，这导致在对检索结果进行性能评价时出现与人类直观感觉不一致的结论。为了克服这个缺陷，提出了一种新的基准数据库分类方案，并且在此基础上提出了一种基于相似性序列分析的检索性能评估方法。

关键词: 三维模型检索, 基准数据库, 相似性序列分析