基于U型脸模型及遗传算法的人脸识别技术研究-计算机软件与理论专业论文.docx

摘 要 本文作者主要做了两方面研究：其一是在人的面部识别算法研究领域首次提出 了 U 型脸的概念，U 型脸是人的面部照片预处理模型，通过实验证明 U 型脸模型在 一定程度上提高了识别率，达到作者的预期效果。其二是利用基因算法进一步改进 了主元分析人的面部识别算法，本文作者在前人的基础上又提出了三点改进，实验 结果表明改进后的算法比前人的算法取得的效果要好。下面详细介绍这两个方面： 1．U 型脸是人的面部照片预处理模型，它把头发、背景等不利于人脸识别的因 素去掉，用标准的背景替代，只留取人脸的关键部位；这样可以在一定程度上消除 时间对人的面部的影响；U-1 是理想模型、概念模型；U-2 是一种类似 U-1 的粗糙 模型；U-3 模型是比较近似 U-1 的模型；U-4 模型是对 U 型脸模型的扩展，该模型 与前三个模型有本质的区别。 2．利用基因算法对主元分析人的面部识别 算法的特征空??进行优化选取，本 文作者在前人的基础上提出了三点改进。传统方法对特征空间的确定有一定的局限 性，并不能达到最优的效果，利用基因算法可以在这方面进行改进。三点改进：首 先是对基因算法编码位数的改进，原来如果是 N 位，现在只需要 N-1 位，并且能够 达到同样的效果，降低了算法的时间复杂度和空间复杂度；其次是前人的初始种群 是随机确定的，本文作者根据特征值及其特征向量的分布规律确定了非随机初始种 群方法，在一定程度上提高了搜索效率；最后是在基因算法运行过程中保存每代最 高适应度的所有染色体，算法运行结束后根据多种方式筛选获得最优结果。 关键词：人脸识别； U 型脸；基因算法；主元分析  Abstract This article focused on two aspects of the research: Firstly, the author first roposed the concept of U-face in the recognition algorithms field, U-face is a face photographs pretreatment model, and simulation results showed that the U-face improved the recognition rate, to the authors desired effect. Secondly, the author further improved PCA facial recognition algorithm by genetic algorithm, the author based on the previous proposed a three-point improvement, experimental results showed that the improved algorithm is better than the previous. These two aspects are detailed below: U-face is the human face photos pretreatment model, it removes background, hair, and other factors that are not conducive to face recognition, and replaced with the standard background, only left the critical parts of the face; to some extent to eliminate that time effects on the human face; U-1 is the ideal model; U-2 is a rough model; U-3 model is more similar to the model U-1; U-4 model is an extension of U-face model, the model is essentially different from the other three models. Genetic algorithm can be applied to optimal selected the feature space of PCA face recognition method， the author based on the previous proposed a three-point improvement. Traditional methods to determine t