面向人脸关键区域的学习算法研究_计算机科学与技术.rar

  • 需要金币1000 个金币
  • 资料目录论文助手 > 高校设计 > 计算机信息 >
  • 转换比率:金钱 X 10=金币数量, 例100元=1000金币
  • 论文格式:Word格式(*.doc)
  • 更新时间:2013-10-07
  • 论文字数:18641
  • 课题出处:(学大教育)提供原创资料
  • 资料包括:完整论文

支付并下载

摘要:随着人工智能理论与技术的发展,计算机视觉技术取得了长足的进步,已经渐渐成为国防建设、公共安全监控、医疗卫生领域应用较为广泛的一项技术。人脸检测与识别作为计算机视觉重要的分支,近年来成为学术界研究的重点,同时也有一些应用项目的开发。

   本文研究和实现了AdaBoost人脸检测算法和基于特征脸的人脸识别算法。结合AdaBoost检测器级联的特性,引入了基于梯度方向金字塔和支持向量机的人脸分类算法,去除了部分误检测的结果;引入了基于肤色模型的分割算法,对人脸区域进行了精确定位。实验在第二代身份证、Feret人脸数据库以及互联网图片资源上进行,给出了实验结果与分析。算法对正面人脸效果显著,但对于旋转和倾斜的人脸,鲁棒性不强。本文最后进行了总结,同时提出了后续研究的方向。

   本文最终的实现在Windows平台下,利用Visual Studio 2008和OpenCV库实现。

关键字:人脸检测;人脸识别;AdaBoost;梯度方向金字塔(GOP);支持向量机(SVM);特征脸

 

Abstract:With the development of theory and technology of artificial intelligence, technologies in computer vision had made considerable progress these years, and were widely applied in the fields of homeland security、video surveillance and medical health. Being an important part of computer vision, the theory and technology of human face detection and recognition had draw great attention in academia world.

   This paper studies and implements the face detection algorithm AdaBoost and face recognition algorithm based on eigenface. Considering the cascade property of AdaBoost algorithm, this paper improves the AdaBoost with a human face classification algorithm based on gradient orientation pyramid and support vector machine. Also an algorithm of image segmentation based on skin color model is introduced to locate the human face accurately. Experiments were carried out on the test set consists of ID-card、Feret database and images from the internet. Experimental results show that both the detection method and recognition method achieve good results on the frontal face images, but less robust on the images of face with apparent rotation and tilt. At last this paper puts forward the future work.

   The implementation and experiment were developed with Visual Studio 2008 on Windows, and OpenCV library were used.

Keywords: Face Detection; Face Recognition; AdaBoost; Gradient Orientation Pyramid; Support Vector Machine; Eigenface

 

本文实现了AdaBoost人脸检测算法,并在这个基础上实现了基于特征脸的人脸识别,对于目前的社会网络监控具有一定的实际意义。主要的应用场景是公共场所视频的监控。对于公安系统对人员的排查和寻找有辅助作用。例如,对网吧、银行、酒店等公共场所的柜台、收银台进行视频监控,可以结合公安系统的逃犯数据库,对公共场所实行视频监控,一旦发现嫌疑人出现在视频中,经过人工确认后,实行抓捕。此外,也可以用于社会人员人连数据库的采集等方面。

 


支付并下载

提示:本站支持手机(IOS,Android)下载论文,如果手机下载不知道存哪或打不开,可以用电脑下载,不会重复扣费