资源描述：
《Key Frame Extraction Based on Artificial Fish》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、2011InternationalConferenceonTransportation,Mechanical,andElectricalEngineering(TMEE)December16-18,Changchun,ChinaKeyFrameExtractionBasedonArtificialFishSwarmAlgorithmandK-meansSunShumin,ZhangJianming,LiuHaiyanCollegeofComputerScienceandCommunicationsEngineeringJiangsuUniver
2、sityZhenjiang,ChinaE-mail:shuminsun0810@126.comAbstract—Keyframeextractionisoneofthemostimportantbutthecalculationisexpensive;(4)theapproachbasedontechnologiesinthecontent-basedvideoretrieval.Inorderto[4-5]clustering.Thekeyframeextractedbyclusteringcanextractkeyframeefficien
3、tlyfromdifferenttypeofvideo,anwellrepresentthemaincontentofthevideo,sothisefficientmethodofkeyframeextractionbasedonimprovedapproachhasbecomethemainstreammethodforkeyArtificialFishSwarmAlgorithmandk-meanswasproposed.frameextraction.However,theapproachneedpredefineFirstly,ani
4、mprovedArtificialFishSwarmAlgorithmwasthenumberofclusterbeforeclustering,andthecomputingappliedtotheextractedcolorfeaturevectortoself-organizedtimeislong.Thesegreatlylimititsfurtherdevelopment.clusterandobtainedaninitialclusteringresult.Secondly,k-ArtificialFishSwarmAlgorith
5、m(AFSA)neednotmeanswasconductedtooptimizetheinitialclusteringresult,designatethenumberofclusterinadvanceandiseasytoandafinalclusteringresultwasobtained.Finally,thecenterfindtheglobaloptimalsolution.Butthespeedofframeofeachclusteringwasextractedasthekeyframe.Asconvergenceisre
6、lativelyslow;K-meansalgorithmisrelevantexperimentshowstherepresentativeofthekeysimple,andtheconvergencespeedisfast.Butitisframeextractedbyusingthisalgorithmarebetterthanotheralgorithmsandtheextractedkeyframecouldsensitivetotheinitialparametersandiseasytofallintoadequatelyexp
7、resstheprimarycontentofthevideo.localoptimization.Asforthementionedproblemsabove,weproposedamethodofkeyframeextractionbasedonKeywords-videoretrieval;keyframe;featureextraction;theimprovedAFSAandk-means.Experimentsshowthatartificialfishswarmalgorithm;k-meansthekeyframeextract
8、edbyusingthisalgorithmhasagoodrepresentative,andcouldadequatelyexpressthepr
