欢迎来到天天文库
浏览记录
ID:34584113
大小:2.37 MB
页数:60页
时间:2019-03-08
《单目多视角显微三维重建关键技术分析》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、浙江工业大学硕士学位论文MULTI-VIEW3DRECONSTRUCTIONUSINGMONOCULARMICROSCOPICIMAGESABSTRACTMicroscopic3Dreconstruction,asacoretechniqueofmicro-visionsystem,hasaveryimportantapplicationvalueinprecisionmanufacturing.However,becauseofobviousmicroscopicimagedegradation,3Dreconstructionfrommulti-viewimagesneedsto
2、beimprovedfrommanyaspectswhenitisappliedinmicrofield.Thegoalofthispaperistoestablishasimpleandefficientmicroscopic3Dreconstructionsystemwithmonocularmicroscopyandapplyittothesurfacereconstructionofmicro-scaleobjects,thusprovidingessentialreferencesformicroscopicinspection.Themaintasksandachieve
3、mentsareasfollows:1.Comparativestudyofimagefocusingevaluationfunction.Imagefocusingevaluationfunctionisoneoftheimportantbasistomakemicro-visionsystemfocusrapidly.Ninespecificimagefocusingevaluationfunctionsofthreekindshavebeenstudied.Howtherobustnessofdifferentkindsofevaluationfunctionchangewhe
4、nmanificationandnoiseintensitychangewasanalyzedthroughexperiments.Thefocusingqualityevaluationschemeofmicroscopicimagecollectionwasfinallydeterminedwiththestudyresults.2.Thestudyofthekeypointdetectionandmatchingalgorithmofimages.Keypointdetectionandmatchingarecrucialelementsofreconstructionprec
5、ision.Thischapterintroducedconstraintbasedonlocalgeometricconsistencyoftransformonthebaseofperformancecomparisonoffourusualkeypointdetectionalgorithmsandthencreatedanewoutlierremovalalgorithm.Thisalgorithmcanobtainmorematchingpointsandmeanwhilereducethematchingerrorsubstantially.Also,itcanelimi
6、natematchingmistakesnearbyepipolarlinewhichcannotbeinspectedbyRANSACandthelike.ii万方数据浙江工业大学硕士学位论文3.Developasetofroundedmicroscopic3Dreconstructionsystem.Themicroscopic3Dreconstructionsystemhardwareinthispaperconsistsof4-DOFmotionplatformandmicroscopicimagingsystem,soitcancollectmulti-viewmicros
7、copicimageseasily.Thenucleusmoduleofthesystemsoftwareintegratedmicroscopicimagecapturingmethod,microscopicimagekeypointdetectionandmatchingalgorithm,andfurthercombinedStructureFromMotionframetoachieveefficientmicroscopic3Dreconstr
此文档下载收益归作者所有