面向逆向工程的工业CT图像矢量化系统改进

面向逆向工程的工业CT图像矢量化系统改进重庆大学硕士学位论文（专业学位）学生姓名：指导教师：段黎明教授兼职导师：刘荣高工学位类别：工程硕士（机械工程领域）重庆大学机械工程学院二O一三年五月ImprovementofindustrialCTimagevectorizationsystemtowardsReverseEngineeringAThesisSubmittedtoChongqingUniversityinPartialFulfillmentoftheRequirementfortheMaster’sDegreeofEngineeringByWangWeiSupervisedby:Prof.DuanLi-mingPluralisticSupervisedbyLiuRongSpecialty:ME（MechanicalEngineeringField）CollegeofMechanicalEngineeringofChongqingUniversity,Chongqing,ChinaMay,2013重庆大学硕士学位论文中文摘要I摘要工业计算机层析技术(简称“工业CT”，IndustrialComputerizedTomography)是逆向工程（ReverseEngineering）领域中一项获取产品数据重要的手段。但是，工业CT技术扫描得到的是二维灰度切片图像，必须进行二维矢量化，才能应用于逆向工程之中。目前已开发的工业CT矢量化系统存在处理效率不高、系统稳定性差和矢量化精度低等问题，影响后续重建工作，本文对该系统进行了改进研究：研究了图像预处理方法：图像增强方法、阈值分割算法和边缘检测方法。通过对比中值滤波和直方图均衡化两种图像增强方法，选择了中值滤波技术作为图像增强方法；根据工业CT图像的不同特征点，选取了迭代法阈值分割和自适应阈值分割两种算法作为图像阈值分割的方法；通过对比经典的边缘检测方法，选取了改进的Canny算子作为系统的边缘检测方法。对特征点检测方法进行了研究，提出了一种满足工业CT图像特征点检测要求的检测方法。首先，通过基于邻域值的轮廓点分类算法对所有轮廓点进行一次筛选获得候选特征点；其次，通过基于曲率的算法定位特征点的位置，减少冗余特征点数量，获得最终的特征点。对圆检测方法进行了研究，提出了改进的点Hough变换圆检测算法。首先通过一次筛选和二次筛选获得候选边缘点组；其次，采用自适应的步长变换值和累加器判断阈值进行圆的检测，获得最终的检测结果。该算法避免了圆检测过程可能出现的误检测和漏检测问题，同时提高了圆的识别和拟合精度。最后，通过与改进前的系统的对比实验，进一步验证了改进工作的有效性。关键词：工业CT，逆向工程，图像预处理，特征点检测，圆检测重庆大学硕士学位论文英文摘要IIABSTRACTThemodelreconstructionwithcomplexfreesurfaceinReverseEngineeringcanofferedbyIndustrialComputerizedTomography.However,the2-dimensionalgrayslicingimagewhichisobtainedbytheICTsystemhavetobeexecutedvectorizationforusingitinCADsoftware.ThedevelopedindustrialCTsystemislowefficiency,poorstabilityandlowaccuracy,whichaffectreconstruction.Sotheimprovementresearchwasrealizedinthispaper:Imagepre-processingmethodswerestudied,includingimageenhancement,thresholdsegmentationandedgedetectionoperator.Medianfiltertechnologywaschoseasimageenhancementmethodthroughcomparisonbetweenmedianfilterandhistogramequalization.AccordingtodifferentindustrialCTimage,iterativethresholdmethodandadaptivethresholdwerebothchoseasimagethresholdsegmentation.ImprovedCannyedgedetectionalgorithmwaschoseasedgedetectionmethodthroughcomparisonamongthetypicaledgedetectionalgorithms.Featurepointdetectionwasresearched,onemethodwasproposedwhichmetpositioningrequirementofindustrialCTimage.Firstly,candidatefeaturepointwasobtainedbasedonneighborhoodsorting;Secondly,alotofredundantfeaturepointswereremovedbycurvaturealgorithm.AnimprovedpointHoughtransformcircledetectionmethodwasproposedbasedoncircledetectionmethodswerestudied.Firstly,candidatepointgroupswereobtainedbytwotimesfilter.Secondly,circledetectionwasexecutedbyadaptivestepvalueandaccumulatorjudgingvalue.Erroneousdetectionandleakdetectionwerebothavoided.Atthesametime,theprecisionofcircledetectionandfittingwasimprovedaswell.Atlast,theimprovedresultwasturnedouttobevalidatedbycomparingwithoriginalsystem.Keywords:IndustrialCT,ReverseEngineering,ImagePre-processing,FeaturePointDetection,CircleDetection重庆大学硕士学位论文目录III目录中文摘要..........................................................................................................................................I英文摘要........................................................................................................................................II1绪论.........................................................................................................................................11.1课题研究背景及意义...............................................................................................................11.1.1工业CT技术概述.............................................................................................................11.1.2逆向工程技术概述............................................................................................................11.1.3研究背景及意义................................................................................................................21.1.4课题来源............................................................................................................................31.2矢量化技术国内外研究现状...................................................................................................31.3本文主要研究内容...................................................................................................................41.4论文的组织结构.......................................................................................................................52工业CT图像矢量化系统简介....................................................................................62.1系统功能模块介绍...................................................................................................................62.1.1图像预处理模块................................................................................................................72.1.2轮廓矢量化模块.................................................................................................................82.1.3数据转换模块.....................................................................................................................82.2系统工作流程介绍...................................................................................................................82.3本章小结.............