智能视频监控中人脸检测的研究与实现  

Research and Realization on Face Detection in Intelligent Video Surveillance

作　　者：谢仪[1,2] 鲍可进[1] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013 [2]江苏科技大学苏州理工学院计算机中心,江苏镇江212004

出　　处：《计算机测量与控制》2013年第11期2921-2923,共3页Computer Measurement &Control

摘　　要：针对智能视频监控中人脸检测的问题,结合运动目标检测和AdaBoost人脸检测算法,提出了一种实时视频中基于局部扫描的快速人脸检测方法;首先利用背景差分法对视频中的运动目标进行检测,标记出运动目标区域作为人脸检测候选区域,然后运用AdaBoost算法对人脸检测候选区域进行人脸检测,返回检测后的结果;对实时视频进行检测,每帧图像平均检测时间为57.8ms,误检率为4.16%;实验结果表明该方法能有效提高基于视频的人脸检测的效率和实时性,同时显著降低了AdaBoost算法的误检率,在一定程度上满足了嵌入式监控平台的应用需求。In order to face detection in intelligent video surveillance, combined with the moving target detection and AdaBoost face detec- tion algorithm, proposed a fast face detection method in real--time video based on partial scan. At first, moving target detection is realized u- sing background difference method in the video, mark the moving target areas as the candidate regions of face detection, then face detection is detected in the areas of face detection candidate regions, return results after detection. To test the real--time video, the average detection time of each video frame is 57.8ms, the false detection rate is 4.16%. Experiments show that this method can effectively improve the effi ciency of video--based face detection and real--time performance, while the false detection rate of AdaBoost algorithm is significantly re duced. To a certain extent, meet the application requirements of embedded monitoring platform.

关 键 词：人脸检测 运动目标检测 ADABOOST算法 背景差分法 智能视频监控 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]