監(jiān)控視頻中特定人查找方法研究.pdf

1、華北電力大學(xué)（保定）碩士學(xué)位論文監(jiān)控視頻中特定人查找方法研究姓名：郭康申請(qǐng)學(xué)位級(jí)別：碩士專(zhuān)業(yè)：信息與通信工程；信號(hào)與信息處理指導(dǎo)教師：戚銀城201203華北電力大學(xué)碩士學(xué)位論文AbstractSpeci6cpeIsonsearchinghasalwaysbeenapopularandalsoachallengingtechnologyinthe矗eldofcomPuterVersionThistechnology’whichcoVers

2、computerVerSion，Videoprocessingandimagesegmentatjon，relatedtomuchtechno】ogyofintelligentVideosurVe訂IanceandVideetrieValtechnologyInsurveillancevideos，SpecincpersonsearChingrequiresthecoInputertorecognizeapersonwithdi行色re

3、ntst8nces，whichisadi蚯cultyofcunentcomputerandVideoprocessingtechnologyCombiningwithlheadVancedimageprocessingandVideoprocessingtechnology，thispaperdiscussesthereIatedtechnoIogyinspecificpeoplesearchingfieldbasedonin—dept

4、hstudyofsuchsystemsThesetechniquesincludeback伊oundmodeling，movingta毽etdetection，andmatchingof佬aturesInmoVingtargetdetection，thispapercomparesseVeralcullrentalgorithms，andaSurendraalgorithmbasedonHSVcolorSpaceisproposedIn

5、鳧aturematching，westudjedglobalfeaturesandlocal先aturesaswenastheirapplicablesituationsEspeciallytheSIFT凳alureisstudiedde叩lyandbeexperimented。SincewecannottakeuseoftheadVantagesofsingle危aturesinspecincpersonsearching，wepro

6、poseamethodofmultifeature如sion1tconsidersthe910balfeature鋤d10cal島ature，whichmaintainsthecorrectnessofsinglefeature，usedincontentretrieval，butalsotakesintoaccounttheimpactofretrievalresultsofdi行宅rentfeaturesThispaperprove

7、sthatthecombined危aturejsbetterthansingle一凳aturewhenusedincontentretrieval1nexperiments，thispaperimplementsaspeci矗cpersonSe雒chingsystemcontainingtwosub—models：moVingta唱etdetectionmodelandfbaturedetectionandmatchingmodelTh

8、eformerisusedtodetectthemovingta唱etsinsurveillancevideoswithbacl‘g即undupdatingalgorithmThelatteristhekeymodelofthissystem，whichisresponSiblebrdetectingtheglobalfea伯feand10calfeature，andthancompletetheret—eValtaskaccordjn