文档介绍：南京气象学院硕士学位论文人脸自动识别系统的研究姓名:程勇申请学位级别:硕士专业:系统分析与集成指导教师:傅德胜 20040501 摘要摘要随着社会的发展、科技的进步,各个方面对快速有效的身份自动验证系统的要求日益迫切。生物特征是人的内在属性,具有很强的自身稳定性和个体差异性,因此生物特征是身份验证系统的理想依据。例如:指纹身份验证系统、瞳孔身份验证系统、语音验证系统等。相对于以上系统的生物特征,人脸特征又具有直接、友好、方便等特点,并易于被用户所接受。由于人脸是非刚体,变化较大,因此人脸识别是一个复杂、涉及面广而且应用前景广阔的研究课题。近几年来,国内外掀起了人脸自动识别的研究热潮,并取得了突破性的进展。本文首先研究了人脸自动识别技术的历史与现状及人脸识别的主要方法,然后提出了一种基于小波变换、主成份分析和人工神经网络的人脸识别方法。本文工作包括: 、人脸预处理,消除了尺度、角度、光照强度差异的影响。 , 建立人脸库。 ,此方法根据人脸图像整体的灰度相关性来提取特征向量,所提取出的特征向量保留了原图像空间中的主要分类信息,并能在均方误差最小的意义上重建原图像。 、改进BP神经网络。 ,并取得了较好的识别性能。本算法同传统的方法相比,可以大大降低训练时间和识别时间,并能显著提高识别率。实验证明本文提出的算法是有效的。关键词:人脸自动识别、小波变换、低频系数、主成份分析、神经网络摘要 ABSTRACT Asthedevelopment of thesociety and theprogress oftechnology,there are increasing demands inautomatic some biological characteristics are stable toPeople and are strongly individual differentfrom one can be used as features instance,fingerprint verification system,pupil verification system,speech verificationsystem withcharacteristicsofabove system,the characteristics of facearemore directly,friendly andconvenient characteristic andCan easily be epted by thecustomers verificationsystem Face isnonrigid and largely variational,SO face recognition is plicated extensive and challenging research there are researchful upsurge intheword andsignificant pmgresses have been made in thetechnology of paper firstlystudy thebackground of face automatic recognition system and methods of face recognition, secondly bring forward a method of face recognition based on wavelet transform,ponent analysis and work. These are the maincontents ofthispaper: First,accurately seeking-pupil,pretreatment forface,avoiding influence ofdifferenceofscale,angle andintensity. Second,using themethod ofwavelet·transform,extract the stable second-low frequency wavelet coefficient,build facedatabase. Third,for the