資源描述：
《基于進(jìn)化算法的三維點(diǎn)云自由拼接研究》由會(huì)員上傳分享，免費(fèi)在線(xiàn)閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、基于進(jìn)化算法的三維點(diǎn)云自由拼接研究StudyofRegistrationfor3DPointCloudsBasedonEvolutionaryAlgorithms(國(guó)家自然科學(xué)基金資助項(xiàng)目:61177002)學(xué)科專(zhuān)業(yè)：光學(xué)工程研究生：周天宇指導(dǎo)教師：葛寶臻教授天津大學(xué)精密儀器與光電子工程學(xué)院二零一五年十一月摘要激光三維掃描技術(shù)是一種快速、高效的三維數(shù)字化手段，目前已經(jīng)廣泛應(yīng)用于形貌測(cè)量、生產(chǎn)制造、逆向工程、影視娛樂(lè)以及人體工程學(xué)設(shè)計(jì)等領(lǐng)域?，F(xiàn)存的三維點(diǎn)云數(shù)據(jù)獲取方法和數(shù)據(jù)處理效率仍然存在上升空間，開(kāi)發(fā)出高效的點(diǎn)云自由拼接算法成為研究的熱點(diǎn)。圍
2、繞這一目標(biāo)，本文開(kāi)展了如下工作：1．進(jìn)行從大量的數(shù)據(jù)點(diǎn)云中快速準(zhǔn)確提取特征點(diǎn)的研究。對(duì)常用的曲率取點(diǎn)、均勻取點(diǎn)、關(guān)鍵點(diǎn)(KPQ)提取、固有形狀特征(ISS)等取點(diǎn)算法進(jìn)行了分析和實(shí)驗(yàn)，討論了不同算法的特點(diǎn)，對(duì)ISS特征點(diǎn)提取算法引入了鄰域半徑約束的改進(jìn)策略，使其廣泛適用于一般性的點(diǎn)云模型的特征提取。2．對(duì)描述兩片點(diǎn)云拼接精度的對(duì)應(yīng)點(diǎn)之間距離中值的目標(biāo)函數(shù)進(jìn)行了分析，論證了使用群智能算法進(jìn)行函數(shù)優(yōu)化的可行性，實(shí)現(xiàn)了粒子群算法(PSO)、生物地理學(xué)優(yōu)化算法(BBO)、人工蜂群算法(ABC)的功能。3.基于特征點(diǎn)提取和群智能優(yōu)化算法，利用Matla
3、b編程實(shí)現(xiàn)了完整的三維點(diǎn)云自由拼接算法，通過(guò)對(duì)理想點(diǎn)云庫(kù)模型和實(shí)測(cè)存在噪聲的點(diǎn)云數(shù)據(jù)的拼接實(shí)驗(yàn)，比較了不同算法對(duì)不同模型的拼接精度和效率。同時(shí)通過(guò)對(duì)不同重合率點(diǎn)云的拼接測(cè)試，獲得了點(diǎn)云重合率與拼接精度的關(guān)系。關(guān)鍵詞：三維點(diǎn)云模型特征點(diǎn)提取算法群智能算法點(diǎn)云拼接ABSTRACTLaser3Dscanningtechnologyhasbeenakindoffastandefficientmeansof3Ddigitalmethod.Ithasbeenwidelyappliedinsurfacetopographymeasurement,garme
4、ntmanufacturing,reverseengineering,filmandtelevisionentertainment,humanbodyengineeringdesignandsoon.Theexistingmethodsof3Dpointcloudsdataacquisitionanddataprocessingefficiencyareexpectedtoreachahigherlevel.Thedevelopmentofefficientpointcloudsregistrationalgorithmshasbecomet
5、hehotspotofacademicresearch.Surroundingontheobjective,themajorworkandinnovationareasfollows:1．Wehavedoneresearchonextractingfeaturepointsinlargepointsquicklyandaccurately.Thispaperanalyzedandcomparedcurvaturefeaturepointsextraction,randomsampling,KPQ-pointsextraction,ISSfea
6、turepointsextraction.WeproposeanimprovedISSalgorithmadoptinganeighbourpointsradiusconstraintstrategy,makeitsuitableforgeneralpointcloudmodels.2．Weanalyzedthefitnessfunctionofthemediandistancebetweencorrespondingpointswhichisusedtodescribetheaccuracyofimageregistrationandpro
7、vedtheswarmintelligencealgorithmcanbeusedtooptimizethefunction.WehaverealizedthefunctionofParticleSwarmOptimization,theBiogeography-basedoptimization,theArtificialBeeColonyAlgorithm.3．Basedontheresearchesabove,thecompleteregistrationalgorithmofpointcloudshasbeendevelopedbyM
8、atlab.Throughtheregistrationresultsofidealmodelsandrealhumanmodel,theperformanceof
