資源描述:
《基于進(jìn)化算法的三維點云自由拼接研究》由會員上傳分享,免費在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、基于進(jìn)化算法的三維點云自由拼接研究StudyofRegistrationfor3DPointCloudsBasedonEvolutionaryAlgorithms(國家自然科學(xué)基金資助項目:61177002)學(xué)科專業(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)用于形貌測量、生產(chǎn)制造、逆向工程、影視娛樂以及人體工程學(xué)設(shè)計等領(lǐng)域?,F(xiàn)存的三維點云數(shù)據(jù)獲取方法和數(shù)據(jù)處理效率仍然存在上升空間,開發(fā)出高效的點云自由拼接算法成為研究的熱點。圍
2、繞這一目標(biāo),本文開展了如下工作:1.進(jìn)行從大量的數(shù)據(jù)點云中快速準(zhǔn)確提取特征點的研究。對常用的曲率取點、均勻取點、關(guān)鍵點(KPQ)提取、固有形狀特征(ISS)等取點算法進(jìn)行了分析和實驗,討論了不同算法的特點,對ISS特征點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云模型的特征提取。2.對描述兩片點云拼接精度的對應(yīng)點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.基于特征點提取和群智能優(yōu)化算法,利用Matla
3、b編程實現(xiàn)了完整的三維點云自由拼接算法,通過對理想點云庫模型和實測存在噪聲的點云數(shù)據(jù)的拼接實驗,比較了不同算法對不同模型的拼接精度和效率。同時通過對不同重合率點云的拼接測試,獲得了點云重合率與拼接精度的關(guān)系。關(guān)鍵詞:三維點云模型特征點提取算法群智能算法點云拼接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