資源描述：
《三維點(diǎn)云數(shù)據(jù)配準(zhǔn)技術(shù)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。
1、大連海事大學(xué)碩士學(xué)位論文三維點(diǎn)云數(shù)據(jù)配準(zhǔn)技術(shù)研究姓名：黃行森申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：馬孜20100601中文摘要摘要三維激光測(cè)量的發(fā)展為實(shí)物的數(shù)字化提供更多的方法。由此也產(chǎn)生了對(duì)來自不同測(cè)量?jī)x器在不同的視覺角度測(cè)定的數(shù)據(jù)匹配的問題。在新興的數(shù)字化維修再制造領(lǐng)域，將實(shí)測(cè)的數(shù)據(jù)與原始的設(shè)計(jì)模型進(jìn)行比對(duì)是不可缺少的技術(shù)環(huán)節(jié)。三維數(shù)據(jù)配準(zhǔn)技術(shù)為解決這些問題提供了有效途徑。本文對(duì)三維數(shù)據(jù)配準(zhǔn)技術(shù)進(jìn)行深入研究，主要研究?jī)?nèi)容及研究成果如下：首先對(duì)數(shù)據(jù)配準(zhǔn)中原始數(shù)據(jù)可能存在噪聲，采取八叉樹結(jié)構(gòu)進(jìn)行去噪，在去除噪聲的同時(shí)，建立良
2、好的數(shù)據(jù)搜索結(jié)構(gòu)。并且為了簡(jiǎn)化后期的配準(zhǔn)過程，基于離散點(diǎn)特征信息對(duì)數(shù)據(jù)進(jìn)行簡(jiǎn)化。同時(shí)為了改進(jìn)初始狀態(tài)不佳時(shí)對(duì)配準(zhǔn)的收斂的影響，采用主成分分析法對(duì)點(diǎn)云數(shù)據(jù)進(jìn)行旋轉(zhuǎn)規(guī)一化。然后針對(duì)最近迭代算法存在對(duì)應(yīng)點(diǎn)選取的問題，借鑒幾何哈希配準(zhǔn)法和擴(kuò)展高斯球法，提出了基于特征信息分段求重心的方法，以此有效減少在尋找對(duì)應(yīng)點(diǎn)對(duì)過程中消耗的時(shí)間。而對(duì)于數(shù)據(jù)配準(zhǔn)中初始旋轉(zhuǎn)矩陣不易確定的情況，結(jié)合遺傳算法的概率搜索獲取初始旋轉(zhuǎn)矩陣，再用最近迭代法進(jìn)行精配準(zhǔn)，二者結(jié)合改善了配準(zhǔn)的效率和精度。最后將配準(zhǔn)算法應(yīng)用于破損修復(fù)系統(tǒng)中，對(duì)破損工件進(jìn)行掃描獲取點(diǎn)云數(shù)據(jù)，將數(shù)據(jù)與設(shè)
3、計(jì)的模型進(jìn)行匹配對(duì)比，基于配準(zhǔn)的結(jié)果提取破損數(shù)據(jù)，為后期的機(jī)器人快速修復(fù)路徑規(guī)劃提供基礎(chǔ)。關(guān)鍵詞：旋轉(zhuǎn)規(guī)一化；三維數(shù)據(jù)配準(zhǔn)；點(diǎn)云信息分類；遺傳算法英文摘要ABSTRACTWiththedevelopmentoflasermeasurementtechnology,morechoicesfor3DsurfacedigitalizingareprovidedSOthatregistrationofdatatechnology，whichalefromdifferentmeasurementequipmentsandviews，isplayin
4、gmoreandmoreimportantrole．Indevelopingdigitalremanufacturingandrepairingfields，itisanimportanttechnologyparttocomparethedatabetweendesignedmodelandmeasuredmodel．Therefore，registrationtechnologyisstudiedandsomeresultsareobtainedasfollows：Firstly,consideringthenoisedatainth
5、eoriginalscanneddata,octreedatastructureisusedtoremovethenoisedataandensuretheaccuracyoffollowingworks．Meanwhile，．inordertOsimplifyandimprovetheefficiencyofregistration,featureinformationofsurfaceisutilized．Astheoriginalpositionofpointcloudmaycausetheunconvergenceofregist
6、ration，aspecialrotationbasedonprimarycomponentanalysisiscalculatedtOregulatetheoriginalposition．Secondly，thekeywaytoimproveICPalgorithmistoadvancetheefficiencyofsearchingthecorrespondingpointsindifferentdatasets．BasedonHashgeometric，Gaussiansphereregisteringalgorithms，gen
7、eticalgorithmanditerativeapproach，anovelregistrationalgorithm，whichgravityiscalculatedinsections，ispresentedSOthattheefficiencyofsearchingthecorrespondingpointsandaccuracyaleimproved．ForGeneticAlgorithmhasagoodperformonprobabilitysearchandoptimization,itisemployedtogainab
8、etterinitialrotationmatrixforiterativeregistration．Finally，toprovidedatabaseforrobotpathplanning
