摘要
1.引言数据库的查询操作一般依赖于物理层的特殊算法的支持,尤其在许多面向对象和空间数据库的应用领域,如在文献数据库、多媒体数据库、金融数据库、CAD数据库等的检索操作中,都需要特殊检索算法支持对象的相似检索。数据库中的对象常常用高维特征矢量表示,因此对象的相似检索问题实际上归结为高维特征矢量的相似检索问题。
This paper introduces a newly generalized structure for the similarity retrieval of high-dimensional feature vectors Called the recursive clustering index tree,which is dynamics,simplicity and suited for various objects. Because the mutually overlapping bundling spheres of cluster may increase the number of search paths, we also propose a similarity retrieval method using pruning branches' strategy. Based on this structure,the retrieval effectiveness of this approach is less than that of the exhaustive search and the similarity retrieval method based on SS-tree,and the retrieval cost of this method is a third of that of the exhaustive search when the recall ratio is less than
出处
《计算机科学》
CSCD
北大核心
1999年第12期62-68,共7页
Computer Science
关键词
数据库
聚类索引树
检索算法
索引结构
R树
High-dimensional feature vectors, Recursive clustering indexing tree, Similarity retrieval method , Pruning branches' strategy