【作者】 吴敏；

【导师】 师黎；

【作者基本信息】 郑州大学 ， 模式识别与智能系统， 2010， 硕士

【摘要】 生物特征识别技术正逐渐成为一种公认的身份认证技术。从最基本的到最完善的,存在着多种不同级别的安全技术,而生物特征识别技术将是最安全的。其中,人脸识别是我们日常生活中最常用的身份认证手段,也是当前最热门的模式识别研究课题之一。总的来说,自动人脸识别系统需要三个步骤：人脸检测与定位,特征选择与提取,人脸的识别。在此背景下,本文设计并实施了一系列针对人脸检测与识别的实验和研究,具体内容如下：首先,本文研究了国内外关于人脸检测和识别的方法,并对这些算法进行了对比和总结,在此基础上确立了本文的研究方向；其次,针对人脸检测系统训练过程的复杂冗长,本文选择研究基于少量训练样本的人脸检测问题。运用基于Fisher判别式分析的线性超平面作为分类器,采用AdaBoost算法构成多层级联分类器进行人脸检测,实验证明本算法减少训练样本却获得了更好的检测效果；最后,本文应用独立分量分析的方法提取出人脸特征,并运用最近邻分类器和支持向量机分别进行了识别,结果表明独立分量分析与支持向量机结合的识别方法效果最佳。更多还原

【Abstract】 Biometric Identification is becoming a recognized identity authentication technology. From the most basic to the most robust, there are different levels of security technology, biometrics will be the safest. The face recognition is the most commonly used in our daily lives means of identity authentication is the most popular research topic in pattern recognition. In generally, automatic face recognition system requires three basic steps to complete:face detection and location, feature selection and extraction, face recognition.In this context, the paper designs and implements a series of face detection and recognition for experimental and research. The following is the main study of this paper.First of all, this paper has studied abroad on the face detection and recognition methods. We compared and summarized these algorithms. Then based on these, we established the direction of this research;Secondly, aimed at the complex and long training process of face detection, this paper address the problem of learning to detect faces from a small set of training samples. Based on Fisher discriminant analysis using linear hyperplane classifier, AdaBoost algorithm constitutes a multi-layer cascaded classifier for face detection. We show that the detection can be significantly improved with our algorithm on a small dataset;Finally, independent component analysis was used to extract the facial features, and the nearest neighbor classifier and support vector machines were both used to identifying, the experiment’s result proves the superiority of independent component analysis and support vector machines.更多还原