基于人脸检测与跟踪智能监控系统_宋红.pdf

24 　 11 V ol. 24　N o. 11 2004 11 Tr ansactions of Beijing Institute of Technolog y Nov . 2004 　　: 1001-0645 2004) 11-0966-05 宋　红,　石　峰 ,　10008 1) 　: ,。 , , ; , , BP ; ,;, ,。、 ,,。 ,、 , 。 : ; ; ; BP : T N 9 11. 7 3　　　: A Face Detection and Tracking for Intelligent Surveillance System SON G Ho ng ,　SHI Feng Depa rtm ent o f Computer Science and Engineering , Schoo l of Info rmation Science a nd Technolog y , Beijing Institute o f Technolog y, Beijing 10008 1, China ) : , Abstract Ai mi ng at ractif ying t he sho rtcomi ng s of t raditio nal v ideo surv eilla nce system an i nt el- lig ent surv eilla nce syst em t ha t can aut omati cally det ect a nd t rack human faces in the scene i s pre- sent ed. Symm etrical f ram e diff erence is used to acquire the area of mo tion and the skin-colo r seg- m enta tion alg orithm based o n BP neural netw o rk is u sed to ex t ract the face ca ndidates. Then, the candidat e face regio ns are v erifi ed w ith t he know ledg e of human f aces. A reco rded face buff er is m ai nt ai ned to track movi ng f aces i n the scene. T he capt ured faces a nd ev ent of int erest are u sed to g enerat e video i ndexi ng data base. Ex perimental resul ts show tha t t he intellig ent surv ei llance system can detect a nd t rack the f aces rapidly a nd accura tely i n the scene. Key words: v ideo surveillance; face det ectio n and t racki ng; sym met rical fram e dif ference; BP neural net wo rk