人脸识别技术设计论文.doc

人脸识别算法
摘要
人脸自动识别是模式识别领域的一项热门研究课题,有着十
分广泛的应用前景。本文对人脸位置矫正,人脸的特征提取和识别这些方面进行了研究,并提出了相应的实现算法。
人脸位置矫正作为人脸检测定位的一个环节,在计算机人脸识别中具有重要的意义。本文第二章提出了一种基于单人脸灰度图像中眼睛定位的人脸位置矫正方法,它是针对人眼灰度变化特点、人眼几何形状特征及双眼的轴对称性而设计的。实验结果表明,该方法对于双眼可见单人脸灰度图像能实现快速有效矫正,并能在矫正结果中精确给出双眼瞳孔位置。
本文第三章提出了一种基于神经网络的主元分析人脸图像识别方法。该方法利用非线性主元分析神经网络对人脸图像提取人脸特征(矢量),并在BP神经网络上实现了对人脸图像的识别。实验结果证明了该方法的有效性和稳定性。
关键词
人脸位置矫正,人脸特征提取,人脸识别,神经网络,灰度图像,图像块纵向复杂度,主元分析法,
The Design and Implementation of Algorithms for Human Face Recognition
Abstract
The automatic recognition of human faces is a hot spot in the field of pattern recognition , which has a wide range of potential applications .
As the results of our in-depth research ,two algorithms are proposed : one for face pose adjustment , the other for facial feature extraction and face identification .
Face pose adjustment , as a loop of human face location, is very important puter face recognition. Chapter 2 of this thesis presents a new approach to automatic face pose adjustment on gray-scale static images with a single face . In a first stage , the right positions of eyes are precisely detected according to several designed parameters which well characterize plex changes of the gray parameter in and around eyes and the geometrical shape of eyes . During the second stage , based on the location and the symmetry feature of eyes , the inclination angle is calculated and the face position is redressed . The experimentation shows that the algorithm performs very well both in terms of rate and of efficiency . What’s more , due to the precise location of eyes , the apples of the eyes are detected .
In chapter 3, a novel approach to human face image recognition based on ponent analysis and works has been proposed . By using BP works , human face images are essfully classified and recognized according to the output of BPNN whose input is the eigenvector extracted from the human face images via nonlinear ponent analysis of a single layer work . Simulation results demonstrate the