基于聚类优化主动模型人脸识别算法研究  

Face Recognition Based on Clustering Optimization Algorithm for Active Model

作　　者：杨新锋[1] 刘平[1] 

机构地区：[1]南阳理工学院计算机科学与技术系,河南南阳473000

出　　处：《计算机仿真》2012年第1期238-241,共4页Computer Simulation

摘　　要：研究人脸识别和跟踪准确度问题。针对在使用大数据样本进行训练前提下,以往AAM算法人脸识别与定位不准确的缺陷,提出了一种新的利用聚类算法对样本空间进行划分,并在此基础上训练多个AAM的分层人脸识别和跟踪算法。首先利用所有训练样本训练得到初始AAM,然后利用一个全新的相似度计算公式,将所有训练样本划分成若干个子类别,在此基础上,针对每个子类别,训练一个相对稳定的AAM。在识别与跟踪过程中,先使用初始AAM进行定位,然后根据子类AAM进行精细化定位,从而得到比以往算法更为精确的定位效果。仿真结果显示改进的算法能准确定位出人脸所在位置,并且具有很高的运算效率,可以方便的实现实时监控系统的人脸跟踪定位及识别等目标。Research face recognition and tracking accuracy. When large data sets are used as training samples, formerly the AAM algorithm for face recognition and positioning was not accurate. The paper put forward a new clustering algorithm to partition the sample space, on the basis of the training of multiple layered AAM face recognition and tracking algorithm. First, all training samples were trained to obtain the initial AAM, and then a new similarity calculation formula was used, and all training samples were divided into several sub categories. On this basis, a relatively stable AAM was trained for each subcategory,. In the recognition and tracking process, the initial AAM positioning was used first, and then according to the class of AAM, the fine positioning was carted out, so as to obtain more accurate positioning effect than the previous algorithms. Simulation results show that the improved algorithm can accurately locate face location, and has the advantages of high efficiency, easy to realize the real-time monitoring system for face tracking and recognition of target positioning.

关 键 词：主动外观模型 聚类分析 人脸识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]