并行二值模式直方图序列人脸识别方法  

Parallel Face Recognition Approach Using Histogram Sequence of Local Binary Patterns

作　　者：杨际祥[1,2] 杨清山[1] 王凡[1] 王荣生[2] 

机构地区：[1]大连理工大学电子与信息工程学院,辽宁大连116024 [2]燕山大学计算机科学与工程系,河北秦皇岛066004

出　　处：《小型微型计算机系统》2010年第4期784-787,共4页Journal of Chinese Computer Systems

基　　金：国家"九七三"重点基础研究发展规划基金项目(2005CB321904)资助;河北省科学研究计划项目(2007492)资助

摘　　要：局部Gabor二值模式直方图序列(histogram sequence of local Gabor binary patterns,简称HSLGBP)的人脸识别方法具有较高的识别率,但该方法的特征计算较复杂、耗时长,并且特征维数高、匹配速度慢.给出一个并行的HSLGBP方法(简称P-HSLGBP),在多核PC机群上使用MPI实现了该方法,并使用该方法对ORL人脸库中的40人共400幅图像做了实验.理论分析和实验说明了P-HSLGBP方法具有较高的加速比和并行计算效率.在保证高识别率前提下,在由10个双核PC机组成的机群环境下的加速比达到17.同时,P-HSLGBP方法具有良好的可扩展性,适于大规模人脸库的快速识别.The method for face recognition,which extracts the histogram sequence of local Gabor binary patterns（HSLGBP） from the magnitudes of Gabor coefficients,can increase the recognition accuracy.However,there are also some problems,such as being time-consuming and low matching rate when dealing with complex and high-dimension feature data.A parallel method for HSLGBP（P-HSLGBP） is presented implemented on Personal Computers（PC） cluster using MPI.The potential of P-HSLGBP is illustrated on the Cambridge Olivetti Research Lab（ORL） face database,which consists of 400 images of 40 individuals.Both theoretical analysis and experiment results indicate that presented method can achieve high speedup ratio and computing efficiency.The speedup and recognition rate for identifying process can be up to 17 and 98.62% respectively in cluster environment consisting of 10 dual-core processor-based PCs.Meanwhile,the presented method is scalable and can also well adapt to large-scale face database.

关 键 词：并行计算 人脸识别 直方图 局部Gabor二值模式(LGBP) 多核PC机群 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]