毕业论文——基于人工神经网络的人脸识别方法的研究与实现.doc

摘 要 人脸识别系统的研究始于上世纪60年代，随着80年代后计算机技术和光学成像技术的发展以及90年代初期模式识别技术的兴起得到了提高，并随着时代的发展逐渐进入了初级应用阶段，尤其是在美国发生911事件以后，引起了各界广泛的重视，自此人脸识别技术在公共安全等领域发挥着愈加重要的作用。与众多生物认证手段相比，人脸识别具有不可代替的显著优点，因而人脸识别逐渐成为当前模式识别和人工智能领域的一个研究热 点。一般来说，人脸识别系统包括图像摄取、人脸定位、图像预处理、以及人脸识别（身份确认或者身份查找）。系统输入一般是一张或者一系列含有未确定身份的人脸图像，以及人脸数据库中的若干已知身份的人脸图像或者相应的编码，而其输出则是一系列相似度得分，表明待识别的人脸的身份。基于人工神经网络的人脸识别的常用学习方法根据网络 结构和模型不同又分为：BP算法和RBF算法。 本文主要讲述的是基于PCA特征提取和BP和RBF神经网络的人脸识别。其中特征提取和特征选择是人脸识别中数据处理阶段的关键技术，PCA特征提取就是将最优的特征子集从原始特征集中选取出来进行某种形式的转换以得到新特征，然后将新特征组合构成特征脸空间，并将其输入经过训练的BP和RBF网络内，比对相似率，以达到人脸识别的目的。 关键词 神经网络 PCA BP RBF ORL 人脸识别 机器学习 ABSTRACT Research on face recognition system began in the 60s of last century, with the rise of development after 80 years of computer technology and optical imaging and 90 time initial stage of pattern recognition technology has been improved, and with the development of the times and gradually into the primary application stage, especially after the 911 incident, attracted widespread attention since then, face recognition technology in the public security field plays a more and more important role. Compared with other biometrics, face recognition has significant advantages can not be replaced,so face recognition has become the current pattern recognition and artificial intelligence. A study on heat Point. Generally speaking, face recognition systemconsists of image acquisition, image preprocessing,face location, and face Recognition ( identification or identity search ). The system input is generally one or a series of containing not determine the identity of the face image, and face database of some known identity face images or the corresponding code, and the output is a series of similarity score, that to identify the face identity. Common learning method for face recognition based on artificial neural network according to the network different can be divided into structural and model: BP algorithm and RBF algorithm. This article is about the PCA feature extraction and face recognition base