海量人脸数据库的快速检索

Method of quick searching in a huge scale face database

针对海量人脸数据库检索时间长的问题,提出了基于L-K均值层次聚类算法。此算法把大型人脸数据库划分成一些子类数据集,对处于类边界的数据,采用冗余技术和预设阈值再重新分配到一些类中,从而使检索过程只在一个或几个子类中进行。实验结果表明,该算法能极大地缩小海量人脸库的检索范围,在保证一定准确率的前提下,有效地提高了检索速度。

An L-K means hierarchy clustering algorithm is proposed to overcome the long time searching in a huge scale face database. By clustering method the whole database is divided into a number of sub-datasets. Data redundant technique and predefined threshold are applied to reassign clustering edge elements of into certain sub-datasets. Then the searching is only carried out in one or few sub-datasets, which greatly reduces the searching time. Experiment results show that the proposed method can significantly reduce the searching range, thus effectively increasing the searching speed while ensuring similar retrieval accuracy as to search the whole database.