基于OpenCV的实时人脸检测算法研究  

Research on Real-Time Face Detection Algorithm Based on OpenCV

作　　者：胡丽清 李启明[1] HU Li-qing;LI Qi-ming(College of Imformation Enginer,ShangHai Maritime University,Shanghai 201306)

机构地区：[1]上海海事大学信息工程学院,上海201306

出　　处：《现代计算机（中旬刊）》2018年第6期44-49,共6页Modern Computer

基　　金：上海海事大学研究生创新基金资助项目(No.2017ycx060)

摘　　要：为了解决传统基于Open CV的人脸检测算法的训练时间和检测时间过长的问题,采用MB_LBP特征代替Haar特征进行提取人脸图像特征并构成人脸检测分类器,并采用图像金字塔法来减少待测图像的数据量和动态肤色提取来限定分类器滑动窗口的滑动范围。实验结果证明,训练分类器时间大大减少,人脸检测的正确率也得到提高,可以达到实时人脸检测。In order to reduce the traditional training time and detection time of face detection algorithm based on OpenCV,proposes a face detection classifier with MB_LBP feature instead of Haar feature.An image pyramid method is used to reduce the data amount of the image to be classified and the dynamic skin color extraction is used to define the sliding range of the classifier sliding window.Experimental results show that the time of training classifier is greatly reduced,the accuracy of face detection is improved,and real-time face detection can be achieved.

关 键 词：MB_LBP特征 图像金字塔 肤色提取 实时检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]