
复杂背景下的人脸检测.doc
31页南京林业大学南京林业大学本科毕业设计(论文)本科毕业设计(论文)题题 目:目: 复杂背景下的人脸检测 学学 院:院: 理 学 院 专专 业:业: 信息与计算科学 学学 号:号: 091101215 学生姓名:学生姓名: 彭 旭 芳 指导教师:指导教师: 李 勇 智 职职 称:称: 教 授 二二 O 一一 三三 年年 五五 月二十二日月二十二日南京林业大学学士论文 1 1摘 要随着人机交互技术日益成为当今研究的一个热点, 人脸检测问题越来越受到重视, 成为模式识别与计算机视觉领域研究的一个热点它在基于内容的图像与视频检索、自动身份识别、智能人机交互、视频监控等领域有着重要的应用价值人脸是一个看似普通其实非常复杂的问题,它所包含的的信息远远多于身体的其他部分,在人脸检测的过程中,经常会出现比较复杂和一些难以预料的情况,例如人脸特征的可变性,背景的复杂性,光线因素等问题的存在都会使得人脸检测存在很大的困难。
本论文围绕复杂背景下静态灰度图像中的人脸检测问题展开了研究,简要介绍了人脸检测现状,重点介绍了检测速度较快,误检率较低,鲁棒性比较好的基于统计理论的AdaBoost算法本文通过借助matlab编程,选取基于积分图的Haar-like 特征,训练弱分类器,级联强分类器最终实现了复杂背景下的人脸检测关键词:AdaBoost Haar-like 特征 积分图 分类器南京林业大学学士论文 2FACE DETECTION IN COMPLEX BACKGROUNDABSTRACTAs human-computer interaction technology is increasingly becoming a hot research today, face detection problem is more and more attention, become a hot research field of pattern recognition and computer vision.It has important applications in the field of content-based image and video retrieval, automatic identification, intelligent human-computer interaction, video surveillance.The face is a seemingly ordinary but actually very complex issue, the information it contains far more than the rest of the body. During the face detection , they would often more complex and unpredictable,such as the existence of the problem of the facial features can be denatured, the complexity of the background, light factors will make the face detection there are great difficulties.This thesis carried out research around the static Gray-scale images in the complex background of face detection, a brief introduction of face detection current situation, focusing on the detection speed, low false detection rate, robustness better based on statistical called AdaBoost ..With Matlab programming, select the integral image-based Haar-like features, training weak classifier cascade strong classifier eventually face detection in complex background.Key Words:AdaBoost ;Haar-like features;Integral image;Classifier南京林业大学学士论文 3目录摘 要 .............................................................................................................................................1 ABSTRACT.....................................................................................................................................2 第一章 绪论 ...................................................................................................................................4 1.1 背景及研究意义 ...............................................................................................................4 1.2 人脸检测技术研究现状 ...................................................................................................5 1.3 人脸检测方法简介 ...........................................................................................................5 1.3.1 基于模板匹配的人脸检测方法 ............................................................................5 1.3.2 基于面部重要特征的人脸检测方法 ....................................................................6 1.3.3 基于统计的人脸检测方法 ....................................................................................6 第二章 人脸检测的评价标 .........................................................................................................8 2.1 检测率 ...............................................................................................................................8 2.2 假阳率 ...............................................................................................................................8 2.3 假阴率 ...............................................................................................................................8 2.4 检测速度 ...........................................................................................................................8 2.5 鲁棒性 ..............................................................................................................................8 第三章 Adaboost 算法的检测过程............................................................................................10 3.1 训练部分 ............................................................10 3.2 检测部分 ............................................................10 第四章 基于 AdaBoost 算法的人脸检测原理及算法实现 ..........................12 4.2 Haar-like 矩形特征 .................................................12 4.2.1 矩形特征的概述 ................................................13 4.2.2 矩形特征值得计算 ..............................................14 4.3 积分图 ..............................................................16 4.4 Adaboost 训练算法........................................................................................................19 4.4.1 训练样本的选择 .................................................................................................19 4.4.2 Haar 特征提取.................................................................。
