人脸识别系统的研究与开发
发布时间：2018-07-13 11:10
【摘要】：人脸识别技术是一种基于人脸面部特征进行身份识别的生物特征识别技术,通过使用摄像头等采集设备提取人脸面部特征,并对其特征进行匹配进而实现身份验证识别。首先,本文介绍了人脸识别技术的产生背景、发展的历史、优势、研究难点以及应用领域,概括总结了人脸识别领域里的一些经典算法并介绍了当前国内外科研机构及公司在人脸识别领域中的重要突破。其次,本文给出了人脸识别的系统构成,并深入研究了各部分所涉及的算法,之后着重分析了人脸特征提取算法对人脸识别性能的影响。人脸识别系统由图像预处理、人脸检测、人脸对齐、特征提取及特征匹配五部分构成。图像预处理部分是通过灰度变换以及直方图均衡化加强图像对比度;人脸检测部分采用Viola-Jones人脸检测算法,通过对基于Harr-like特征的训练集进行训练得到Adaboost强分类器进行人脸分类,并通过级联分类器结构提高检测的速度;人脸对齐部分依据双眼坐标进行平面几何变换实现人脸的标准化;特征提取部分深入研究了局部二值模式算法(LBP)和局部相位量化算法(LPQ),其中LBP提取的是局部区域像素差异信息,计算简便高效,而LPQ提取的是局部相位信息,对于模糊图像有着很强的鲁棒性;特征匹配部分使用卡方距离计算LBP、LPQ算法提取的样本空间距离,并研究了接收阈值的设置原理。通过对LBP、LPQ算法的仿真分析,评估这两种算法的差异以及在不同的人脸数据库平台下不同条件的识别率。最后,本文人脸识别系统的实现是基于Qt可视化开发平台,调用计算机视觉库Open CV中相关函数模块,尤其是人脸检测模块,方便快捷的实现系统的开发工作。文中同时也给出了Qt、Open CV的相关环境参数设置及编译方法、系统用户界面、功能模块以及系统最终的输出结果。
[Abstract]:Face recognition technology is a biometric recognition technology based on facial features. Face features are extracted by using camera and other acquisition devices, and their features are matched to achieve identity identification. First of all, this paper introduces the background, development history, advantages, research difficulties and application fields of face recognition technology. This paper summarizes some classical algorithms in the field of face recognition and introduces some important breakthroughs in the field of face recognition made by domestic and foreign scientific research institutions and companies. Secondly, this paper gives the system structure of face recognition, and deeply studies the algorithms involved in each part, and then analyzes the influence of face feature extraction algorithm on the performance of face recognition. Face recognition system consists of five parts: image preprocessing, face detection, face alignment, feature extraction and feature matching. In the image preprocessing part, image contrast is enhanced by gray level transformation and histogram equalization. In the face detection part, the Adaboost strong classifier is obtained by training the Harr-like training set by using Viola-Jones face detection algorithm. The speed of detection is improved by cascaded classifier structure, and face alignment is realized by plane geometry transformation according to binocular coordinates. In the feature extraction part, the local binary mode algorithm (LBP) and the local phase quantization algorithm (LPQ) are studied in depth. The local pixel difference information extracted by LBP is simple and efficient, while the local phase information is extracted by LPQ. In the feature matching part, chi-square distance is used to calculate the sample space distance extracted by LBPU LPQ algorithm, and the principle of setting reception threshold is studied. The difference between the two algorithms and the recognition rate of different conditions under different face database platforms are evaluated by the simulation analysis of LBP- LPQ algorithm. Finally, the realization of the face recognition system is based on the QT visual development platform, which calls the correlation function module in the computer vision library Open CV, especially the face detection module, and realizes the development of the system conveniently and quickly. At the same time, the environment parameter setting and compiling method, the user interface, the function module and the final output result of QtnOpen CV are also given in this paper.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41
本文编号：2119154
[Abstract]:Face recognition technology is a biometric recognition technology based on facial features. Face features are extracted by using camera and other acquisition devices, and their features are matched to achieve identity identification. First of all, this paper introduces the background, development history, advantages, research difficulties and application fields of face recognition technology. This paper summarizes some classical algorithms in the field of face recognition and introduces some important breakthroughs in the field of face recognition made by domestic and foreign scientific research institutions and companies. Secondly, this paper gives the system structure of face recognition, and deeply studies the algorithms involved in each part, and then analyzes the influence of face feature extraction algorithm on the performance of face recognition. Face recognition system consists of five parts: image preprocessing, face detection, face alignment, feature extraction and feature matching. In the image preprocessing part, image contrast is enhanced by gray level transformation and histogram equalization. In the face detection part, the Adaboost strong classifier is obtained by training the Harr-like training set by using Viola-Jones face detection algorithm. The speed of detection is improved by cascaded classifier structure, and face alignment is realized by plane geometry transformation according to binocular coordinates. In the feature extraction part, the local binary mode algorithm (LBP) and the local phase quantization algorithm (LPQ) are studied in depth. The local pixel difference information extracted by LBP is simple and efficient, while the local phase information is extracted by LPQ. In the feature matching part, chi-square distance is used to calculate the sample space distance extracted by LBPU LPQ algorithm, and the principle of setting reception threshold is studied. The difference between the two algorithms and the recognition rate of different conditions under different face database platforms are evaluated by the simulation analysis of LBP- LPQ algorithm. Finally, the realization of the face recognition system is based on the QT visual development platform, which calls the correlation function module in the computer vision library Open CV, especially the face detection module, and realizes the development of the system conveniently and quickly. At the same time, the environment parameter setting and compiling method, the user interface, the function module and the final output result of QtnOpen CV are also given in this paper.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41
【参考文献】
相关期刊论文 前2条
1 甘玲;朱江;苗东;;扩展Haar特征检测人眼的方法[J];电子科技大学学报;2010年02期
2 肖冰,王映辉;人脸识别研究综述[J];计算机应用研究;2005年08期
相关硕士学位论文 前3条
1 卜秋月;基于OpenCV的人脸识别系统的设计与实现[D];吉林大学;2015年
2 孙志;基于OpenCV的人脸识别算法实验平台研究与实现[D];吉林大学;2014年
3 王英明;基于AdaBoost算法的人脸检测研究[D];南京理工大学;2008年
，本文编号：2119154
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2119154.html
