基于相似性聚类的多媒体数据库检索算法  

Multimedia database retrieval algorithm based on similarity clustering

作　　者：郑芳霞 Zheng Fangxia(Department of Science and Technology,Zhengzhou Preschool Education College,Zhengzhou 450000,China)

机构地区：[1]郑州幼儿师范高等专科学校科学与技术学院,郑州450000

出　　处：《现代计算机》2023年第23期32-36,共5页Modern Computer

摘　　要：针对现有检索方法对多媒体数据库检索时,多是以查询内容特征为索引,存在查准率和查全率过低,影响检索综合有效性的问题,引入相似性聚类技术,开展多媒体数据库检索方法设计研究。通过借助聚类方法对数据库中的数据进行分区,生成相似数据对象的聚类集合,并完成数据放置。针对不同类型的聚类集合,采用不同的处理方式来获得各自的特征,使用余弦距离方法来比较不同类别的特征向量之间的相似性,并对特征向量进行降维处理。将降维后的特征表示和相似度度量作为示例特征,应用动态分区分配算法实现多媒体数据库检索。通过对比实验证明:新的检索算法具备更高的查准率和查全率,检索的综合有效性更强,应用价值更高。In response to the existing retrieval methods for multimedia databases,which mostly rely on query content features as the index,there is a problem of low precision and recall,which affects the overall effectiveness of retrieval.Therefore,similarity clustering technology is introduced to carry out research on the design of multimedia database retrieval methods.By using cluster-ing methods to partition data in the database,a clustering set of similar data objects is generated,and data placement is completed.For different types of clustering sets,different processing methods are used to obtain their respective features.The cosine distance method is used to compare the similarity between feature vectors of different categories,and the feature vectors are dimensionally reduced.Using the dimensionality reduced feature representation and similarity measurement as example features,dynamic parti-tion allocation algorithm is applied to achieve multimedia database retrieval.Through comparative experiments,it has been proven that the new retrieval algorithm has higher precision and recall,stronger comprehensive effectiveness of retrieval,and higher appli-cation value.

关 键 词：内容 数据库 提取 特征 检索 多媒体 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP391.3[自动化与计算机技术—计算机科学与技术]