面向人機(jī)交互的人臉識別方法研究.pdf

1、大連理工大學(xué)碩士學(xué)位論文面向人機(jī)交互的人臉識別方法研究姓名：趙淼申請學(xué)位級別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：王偉20051201趙淼：面向人機(jī)交互的人臉識別方法研究ResearchonFaceRecognitionforHuman—RobotInteractionAbstractForthewidelyuseofmobile—robotinservicearea，peoplepaymoreandmoreattentiontoth

2、efriendlymethodofhumanrobotinteraction，likefacerecognitionThisthesisresearchesthecoretechnologiesoffacerecognitionsuchasthetechnologyofface10cation，facefeaturerecognitionandtracking，whichcaneffectivelyoperateonmobilerobo

3、tinrealtimeinvariousillumirmtionandfaciaigesturenlethesissummarizestheresearchanddevelopmentofhuman—robotinteractionandfacerecognitionInordertotracktherotationoffacecenter，skincolordetectionandimagedifferencearepresented

4、tolocatefaceundercomplicatedbackground，bilateralsymmetryofgradientdirectionandgeometricrelationshipoffacefeaturesareusedtoextractmidlineoffaceandpositionoffacefeattires，thenthecenterofeyesisobtained，whichnearlyrepresents

5、thecenterofthefacefortrackingTlismethodcallsolvethetrackingproblemduetofacefeatures’shapechangecausedbyfacerotationincertainanglerangeTosolvetheproblemofthetrackingoffacialgesture，skincolordetectionandellipseapproximatio

6、nateusedtolocateface；gradientprojectionofspecialareaisappliedtoextractthefeaturesininitialframethenKalmanfilterisadoptedtoforecastthepositionoffeaturesinfoltowedflameswhichisrenewedbygradientprojectioninsmallfeaturewindo

7、wsTakingtheaveragepositionoffeaturesasfacecenterfortrackingthemethodrealizesthecomprehensionofsimplefacialgestureThismethodcantrackfacefeaturesinbothhorizontalandverticalrotationangle，itcanbeusedforanykindofhumanandinmor

8、econditioncomparedtotemplatematchExperimentalresultsandfurtherexperimentaldataanalysisshowthemethod’svalidityandpracticabilityKeyWords：FaceRecognition；HumanRobotInteraction；BilateralSymmetry；GradientProjection：KalmanFilt