设计说明书 智能安防人脸识别系统研究与开发.pdf

中文摘要 本课题目标是要设计一个智能安防人脸识别系统。系统可以进行图像获取及 检测，实时实现被测人的身份验证，故系统应该具备基于视频流的人脸图像实时 获取、实时检测的功能。为此研究了识别过程中用到的主要算法，进行了验证。 本文主要利用图像灰度化、图像几何归一化、直方图均衡化方法对视频图像 进行预处理。通过对人脸检测中 Haar 特征以及 Adaboost 算法的应用研究，结合 了人脸分类器以及人眼分类器的使用来进行人脸检测，降低误检率。本文提出了 一种基于 LBP 算子和主成份分析算法相结合的人脸图像识别算法，利用 LBP 算 子提取人脸纹理特征的高性能以及主成份分析算法降维以及识别，能有效降低高 维数据的特性，提高识别率。实验表明，与其它算法比较，此算法能够对人脸 像进行良好的分类识别。在系统设计方面，利用 Visual C++ 6.0 开发平台和 OpenCV （OpenCV Source Computer Vision Library ，开源的计算机视觉库）设计 了智能安防人脸系统软件，此软件进一步验证了算法的有效性。 关键词 视频图像，人脸检测，人脸识别 外文摘要 Title Research and Development for Face Recognition System of Intelligent Security Abstract The aim of the subject is to design a system of intelligent security which can recognize face accurately. The image can be got and detected by the system to recognize the identity of the detected person in real time. The system should have the function of face image acquisition, real-time tracking, real-time detection based on streaming video. So, the main algorithms used in the identification process are studied and validated ,the main research in this paper are as follows: In this thesis, basic methods, which are image grayscale, image geometric normalization and histogram equalization, are used to pretreat the video image. The application of face detection in Haar features and the Adaboost algorithm are studied. The face classifier and the eye classifier are combined to detect face, which can reduce the false detection rate. The human face recognition algorithm based on the combination of the LBP operator and PCA algorithm is presented in this paper. The method of LBP is used to extract facial texture features in a high performance and the method of PCA to reduce the dimension and to recognize, which can effectively reduce the char