基于LBP和OpenCV的实时人脸检测系统设计

为了在实时的视频流中快速准确地检测人脸信息,本文采用了一种基于人脸局部LBP特征的检测方法,利用Open CV视觉开发库,在Visual Studio2010环境下采用C++语言开发了一套实时人脸检测系统,利用Open CV中的两个工具opencv_traincascade和opencv_createsamples对样本数据库AR和FERET中的图像进行学习训练,通过训练后的人脸和人眼分类器face.xml和eye.xml文件对实时图像进行人脸检测。测试表明,相较于传统采用Haar特征的人脸检测系统,本文采用的LBP特征人脸检测系统,在检测速度和准确性方面效果更好,本系统也可以在移动平台上部署。

Design of real time face detection system based on LBP and OpenCV

In order to detect fast and precisely the face information in real time video stream,a method based on the local LBP features of the face detection is presented,utilizing OpenCV visual development library,in the Studio2010 Visual environment using C++ language and developed a real-time face detection system. OpenCV’s two tools,opencv_traincascade and opencv_createsamples,are used in the system to train the sam-ple AR database and FERET images. The trained face and eye classifier face.xml and eye.xml file are ob-tained,and finally the faces are detected on real-time image through the two classifiers. Tests show that,com-pared with traditional haar feature based system,our system based on LBP feature outperforms in detecting speed and accuracy. It can also be applied on mobile devices.