資源描述：
《基于光流的圖像目標(biāo)跟蹤方法研究》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、摘要?jiǎng)討B(tài)序列圖像跟蹤是計(jì)算機(jī)視覺(jué)研究領(lǐng)域中十分重要的問(wèn)題。光流法在序列圖像的運(yùn)動(dòng)轂標(biāo)捻測(cè)與跟蹤中褥舞了較好躲應(yīng)用，光渡法用于運(yùn)動(dòng)爨標(biāo)趿蹤存在的問(wèn)題是；運(yùn)算量大、抗噪性蓑、不能跟蹤較贏速度的運(yùn)動(dòng)墾挺：特縫光浚法可較好f}鏨瓣決上述l、蠢題。本文對(duì)光浚法建予運(yùn)動(dòng)爨標(biāo)躐黥避行了磺突，主要工作鰩下：1．對(duì)攀予鑿率的角點(diǎn)檢溺算法和Plessy角點(diǎn)檢測(cè)算法做了理論分析和實(shí)驗(yàn)琵較，通過(guò)對(duì)角煮稔測(cè)算法靜三個(gè)瞧能指標(biāo)(穩(wěn)定襤、可靠性和抗嗓性)酌定量評(píng)價(jià)表明，基于曲率的角點(diǎn)檢測(cè)算法比Plessy角點(diǎn)檢測(cè)算法性能更好。2．研究了經(jīng)典基于微分的光流計(jì)算方法
2、茅口五點(diǎn)約束光流算法，繪出了魏弛算法用于運(yùn)動(dòng)目標(biāo)提取的仿輿實(shí)驗(yàn)比較。實(shí)驗(yàn)結(jié)果表明：五點(diǎn)約束光流箕法具有糖發(fā)毫，計(jì)算速度姨的特點(diǎn)。3。將五點(diǎn)約束光淡算法皮用于不同的運(yùn)動(dòng)目標(biāo)嬲像跟蹤，仿褰結(jié)果表明該方法在跟蹤運(yùn)動(dòng)葦鑭辯，靈縫跟蹤囂櫟大致輪瘁，毽不能給掰完熬讖藩。4．磷究了特，霞光滾簿予運(yùn)動(dòng)瓣標(biāo)圖像跟蹤靜溺越，針對(duì)鞋往運(yùn)動(dòng)昏際實(shí)時(shí)踉蹤圈難和經(jīng)歷旋轉(zhuǎn)時(shí)的失躐問(wèn)題，給出了光流聚類規(guī)則，礙f入矩特征進(jìn)行矩特征蔽配，幽就提島了基于澈特征和特征光流的運(yùn)動(dòng)目標(biāo)圖像鼴稼方法。5．詳細(xì)給出了基于矩特缸和特征光流的運(yùn)動(dòng)目標(biāo)跟蹤方法的算沒(méi)流稷，并進(jìn)行了圖像舅
3、標(biāo)跟蹤仿真。仿真結(jié)果襲明本文提出黢圖像跟踩算法在跟蹤速度，跟蹤精度．和克服旋轉(zhuǎn)三方顢性能優(yōu)越，同時(shí)目標(biāo)做30度以內(nèi)的旋轉(zhuǎn)時(shí)，能夠穩(wěn)定數(shù)跟爨遴動(dòng)娶標(biāo)。關(guān)鍵蠲；踅像痔剜、光流法、特短光流、運(yùn)渤舀稱跟蹤、矩特征阿北T業(yè)人學(xué)塒
4、f學(xué)位論文ABSTRACTThesequenceimagetrackingisanextremelyimportantprobleminthefieldofcomputervision．Theopticalflowmethodshavebeenputintoabetterapplicationindetectiona
5、ndtrackingofmovingobjectsforsequenceimage．Butthemethodshavethefollowingfaults：heavycomputationalburdenandweakanti—noiseability．Furthermore，itcannottrackthemovingobjectswithhighspeed，了kfeature·optical—flowmethodhasthebettersolutionsforthefaults．Thethesisresearchesthefea
6、ture*opticalflowmethodOilthetrackingmovingobjects；themainworksareasfollows：1．Thetheoreticalanalysisandtheexperimentalcomparisonaregivenforcornerdetectionbasedonthecurvaturealgorithmand攮ePlessycomerdetectionalgorithm，F(xiàn)orthetwocornerdetectionalgorithms，threequantitativee
7、valuations(stability,reliabilityandanti-noise)showthattheperformanceofcomerdetectionalgorithmbasedonthecurvatureiSbetterthanthePlessycomerdetection．2。TheclassicalopticalflowmethodsbasedOndifferencetheoryandbasedonfive—pointrestraintsareresearched．Theperformancecomparis
8、onsofthetwoalgorithmsonextractionofmovingobjectsaregiven．Thesimulationresultsconfirmthatfive-pointrestraintsopticalflowalgorithmhasthebetterprecisionwithlesscomputationalburden．3．Thefive‘pointrestraintsopticalflowalgorithmisappliedtotrackingmovingobjects．Simulationresu
9、ltsdemonstratethatwhentrackingmovingvehiclesthisalgorithmcanonlyobtaintheoutlinesinsteadofthewholepicture．4．Theproble
