hmm based detection of fire, falling person and:基于hmm的检测火灾,落人

hmm based detection of fire, falling person and:基于hmm的检测火灾,落人

ID:33117174

大小:758.50 KB

页数:30页

时间:2019-02-20

hmm based detection of fire, falling person and:基于hmm的检测火灾,落人_第1页
hmm based detection of fire, falling person and:基于hmm的检测火灾,落人_第2页
hmm based detection of fire, falling person and:基于hmm的检测火灾,落人_第3页
hmm based detection of fire, falling person and:基于hmm的检测火灾,落人_第4页
hmm based detection of fire, falling person and:基于hmm的检测火灾,落人_第5页
资源描述:

《hmm based detection of fire, falling person and:基于hmm的检测火灾,落人》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、Deliverable11.4REPORTONPROGRESSWITHRESPECTTOPARTIALSOLUTIONSONHUMANDETECTIONALGORITHMS,HUMANACTIVITYANALYSISMETHODS,ANDMULTIMEDIADATABASESPreparedbyA.EnisCetinBilkentUniversityAnkara,TurkeySummary:Thisreportconsistsoftwoparts.Thefirstpartoftheprojectdescribesthepart

2、ialsolutionsdevelopedbythemembersofWP-11onhumanactivitydetectionandunderstandingsystemsinvideointhelasteighteenmonths.InthesesystemsfeaturesextractedfromvideoandassociatedaudioareprocessedbyHiddenMarkovModels(HMMs),NeuralNetworks(NN),DynamicProgramming(DP),andeigens

3、pacemethodstoextractmeaningfulinformation.Inthisreport,thefollowingmethodsaredescribed:-HumanfacedetectioninvideousingDP,-humanbodydetectionusingclassificationofmovingregionsinvideo,-movingobjectclassificationbyeigenanalysisofperiodicmotion,-detectionoffallingperson

4、sinvideousingbothaudioandvideotrackdata.-detectionoffireinvideo,and-detectionofmovingtreeleavesinvideo.Inthesecondpartofthereport,ourresearchonmultimediadatabasesandthepartialsolutionstothefollowingproblemsaredescribed:-Naturallanguageinterfacesforinteractionwithmul

5、timediadatabases,-gesture-basedinteractionwithmultimediadatabases,and-textrecognitioninvideo.Finallygapsinknow-howandrecommendedfuturedirectionsarediscussed.1.HumanFaceDetectioninVideoOurcurrentresearchisbasedontheideathatafacecanberecognizedbyitsedges.Infact,acaric

6、aturistdrawsafaceimageinafewstrokesbydrawingthemajoredgesoftheface.Mostwaveletdomainimageclassificationmethodsarealsobasedonthisfactbecausewaveletcoefficientsarecloselyrelatedwithedges,seee.g.,[Garcia1999].Thefirststepofthealgorithmistofindpossiblefaceregionsinatypi

7、calimageorvideo.Thiscanbedeterminedbydetectingregionswithpossiblehumanskincolors.Intheseregionsedgesareestimatedbyusingastandardedgedetector.Edgescanbeestimatedbysummingtheabsolutevaluesofhigh-low,low-highandhigh-highwaveletsub-imagesofthetwo-dimensionalwavelettrans

8、form.Waveletcoefficientsofthelow-high(high-low)sub-imagecorrespondtohorizontal(vertical)edgesoftheimageregion.Aftercomputingtheedgesinskin

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。