[1]李士进,杨静宇,金忠.基于离散隐马尔可夫模型和奇异值特征的人脸检测[J].南京理工大学学报(自然科学版),2000,(05):385-389.
 LiShijin YangJingyu JinZhong.Face Detection Based on Discrete Hidden Markov Models and Singular Value Feature[J].Journal of Nanjing University of Science and Technology,2000,(05):385-389.
点击复制

基于离散隐马尔可夫模型和奇异值特征的人脸检测()
分享到:

《南京理工大学学报》(自然科学版)[ISSN:1005-9830/CN:32-1397/N]

卷:
期数:
2000年05期
页码:
385-389
栏目:
出版日期:
2000-10-30

文章信息/Info

Title:
Face Detection Based on Discrete Hidden Markov Models and Singular Value Feature
作者:
李士进杨静宇金忠
南京理工大学计算机系, 南京210094
Author(s):
LiShijin YangJingyu JinZhong
Department of Computer,NUST,Nanjing 210094
关键词:
图象处理 建立模型 检测 模式识别 人脸 隐马尔可夫模型 奇异值特征
Keywords:
imag e processing model building measuring pat tern recognit ion face hidden Markov models( HMM) singular value feature
摘要:
该文提出了基于离散隐马尔可夫模型 (HMM )和奇异值特征的人脸检测方法。这一算法包含 2部分工作。首先 ,提出了基于离散隐马尔可夫模型和奇异值特征的正面端正人脸检测方法 ;然后将该算法扩展到检测任意旋转角度的人脸。扩展算法首先计算当前位置子图像窗口的奇异值特征 ,将该特征向量经过识别各个旋转角度人脸的HMM模型 ,得到该子图像窗口的旋转角度 ,再经过旋正 ,重新送到识别正面端正人脸的HMM模型 ,由此确定该子图像窗口是否为人脸。对一个由 43幅集体照片组成的正面人脸图像集进行测试 ,共检测到 484人中的 42 5人 ,检测率为 87.8% ;而多角度旋转人脸图像检测率为 75 .1%。实验结果表明 ,该方法具有良好的检测性能
Abstract:
T his paper presents a novel face detect ion method w hich consist s of tw o algorithms. The f irst one is a f ronta-l view upright face detection algorithm based on the w el-l know n singular value feature ( SVF) and discrete hidden Markov models ( HMM) . The algorithm couples the v irtues of both the SVF and HMM and produces excellent detect ion results. It is tested on a collect ive photo album and has detected the 87. 81 percent of it s 484 people, w hile 537 false alarms are reported. The second one is the extension of the f irst one to rotat ioninvariant face detection. Several HMMs are ut ilized simultaneously and the angle of the ’face’image is obtained. Then the HMM for detecting the upright faces is employed to verify the faceness of the test pat tern. This rotat ion- invariant algorithm is tested on another image set of 173 persons and the detect ion rate is 75. 14 %.

参考文献/References:

1 Colmenarez A J, Huang T S. Face detect ion w ith information- based maximum discrimination.Puerto Rico: IEEE Conference on Computer Vision and Pattern Recognition, 1997. 782~ 787
2 Sung K K. Learning and example select ion for object and patter n detect ion: [ Ph. D. dissertaion] .Boston: M IT Artificial Intelligence Laboratory , 1996
3 Yang G Z, Huang T S. Human face detection in complex background. Pattern Recognition, 1994,27( 1) : 53~ 63
4 Rowley H A, Baluja S, Kanade T. Neural netwo rk- based face detection. IEEE Trans Patter n Analysis and Machine I ntelligence, 1998, 20( 1) : 23~ 38
5 姜志华, 郭跃飞. 基于灰度突变特征结构的快速人脸检测. 南京理工大学学报( 自然科学版) , 2000, 24( 2) : 143~ 146
6 洪子泉, 杨静宇.用于图像识别的图像代数特征抽取. 自动化学报, 1992, 18( 2) : 232~ 238
7 Rabiner L R. A tutorial on hidden Mar kov models and selected applications in speech r ecognition.Proceedings of the IEEE, 1989, 77( 2) : 257~ 286
8 李士进.人脸检测与识别方法研究: [ 博士学位论文] . 南京: 南京理工大学, 2000
9 金忠.人脸图像特征抽取与维数研究: [ 博士学位论文] . 南京: 南京理工大学, 1999
10 蔡元龙.模式识别.西安: 西北电讯工程学院出版社, 1986

相似文献/References:

[1]关英姿,杨大明.航天器对接过程的动力学建模[J].南京理工大学学报(自然科学版),2001,(01):6.
 GuanYingzi YangDaming.Dynamic Modeling of Spacecraft Docking Process[J].Journal of Nanjing University of Science and Technology,2001,(05):6.
[2]丁玉庆.汽车振动系统的简化及数学模型的建立[J].南京理工大学学报(自然科学版),2001,(04):391.
 DingYuqing.On Setting up Simplified and Mathematical Model of Truck Vibration System[J].Journal of Nanjing University of Science and Technology,2001,(05):391.
[3]姜志华,郭跃飞,娄震,等.基于灰度突变特征结构的快速人脸检测[J].南京理工大学学报(自然科学版),2000,(02):143.
 JiangZhihua GuoYuefei LouZhen YangJingyu.Fast Human Face Detection Based on Grey Discontinuous Feature Structure[J].Journal of Nanjing University of Science and Technology,2000,(05):143.
[4]李宏东,李宏升,顾伟康.应用专门知识引导的联机自然手写汉字识别[J].南京理工大学学报(自然科学版),2000,(02):147.
 LiHongdong LiHongsheng GuWeikang.A Knowledge Based Algorithm for Online Cursive Chinese Character Recognition[J].Journal of Nanjing University of Science and Technology,2000,(05):147.
[5]董涛,李迎,侯丽雅.磨削温度在线监测与预报系统的建模研究[J].南京理工大学学报(自然科学版),2000,(05):398.
 DongTao LiYing HouLiya.Modeling of Real-time Monitoring and Forecasting System about Grinding Temperature[J].Journal of Nanjing University of Science and Technology,2000,(05):398.
[6]陈进,周心广,徐骏善,等.信息模型方法在建立FMS数据库中的应用[J].南京理工大学学报(自然科学版),1999,(03):13.
 ChenJin ZhouSongQijun ZhangYouliang.Application of Information Model in FMS Database Building[J].Journal of Nanjing University of Science and Technology,1999,(05):13.
[7]陆建峰,李士进,唐振民,等.基于遗传算法的二维熵方法自动阈值[J].南京理工大学学报(自然科学版),1998,(02):9.
 Lu Jianfeng L i ShijinT ang ZhenminYang Jing yu.The Automatic Thresholding Using Two dimensional Entropy Method with Genetic Algorithms[J].Journal of Nanjing University of Science and Technology,1998,(05):9.
[8]王加存,姜胜兵,王志刚.指挥控制系统信息处理容量模型[J].南京理工大学学报(自然科学版),1997,(05):15.
 WangZhigang.A Model for Information Processing Capacity of Command and Control Systems[J].Journal of Nanjing University of Science and Technology,1997,(05):15.
[9]徐明友,丁松滨,徐直军.外弹道优化设计的随机模型[J].南京理工大学学报(自然科学版),1997,(05):23.
 XuMingyou DingSongbin XuZhijun.Random Model for Optimum Design of Exterior Ballistics[J].Journal of Nanjing University of Science and Technology,1997,(05):23.
[10]赵建,王丽芳,符意德.Petri网在MIS结构设计中的应用[J].南京理工大学学报(自然科学版),1996,(03):257.
 Zhao Jian,Wang Lifang,Fu Yide.Application of Petri Net in MIS Structure Design[J].Journal of Nanjing University of Science and Technology,1996,(05):257.

备注/Memo

备注/Memo:
国家自然科学基金资助项目
李士进 男 27 岁 博士生
更新日期/Last Update: 2013-03-25