符号结构光图像解码及重建曲面滤波方法的研究

《符号结构光图像解码及重建曲面滤波方法的研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、浙江理工大学硕士学位论文符号结构光图像解码及重建曲面滤波方法的研究姓名：李华强申请学位级别：硕士专业：计算机应用技术指导教师：喻擎苍20080301对结构光投射模式中信号点的分布密度，采用多阈值对结构光图像进行分辨识别，并通过桥接点的模糊聚类对多分辨识别结果进行覆贴，大大提高了符号的识别率。(5)基于曲面拟合的滤波方法研究。针对三维重建曲面中存在的噪音，研究了基于移动小区域三次B样条曲面拟合的曲面滤波方法，有效降低了拟合曲面中的噪音，改善了重建结果。关键词：结构光；符号；解码；曲面；滤波ⅡAbstractComp

2、mervisionhasbeentheinvestigativehotspotforseveraldecadesandhasbeenappliedinmanyindustrialandagriculturalfieldssuchasrobotnavigation,objectrecognition，industrialautomationandtargettracking，etc．．Structuredlighttechniqueisabranchofbinocularcomputervisionbyutiliz

3、ingaprojectorinsteadofonecamerainbinocularstereovision．Becausetheprojectorstructuredlightpattemownsthecharacteristicofcodeuniqueness。structuredlighttechniqueisabletosolvethepixelmatchingprobleminbinocularvision．andiSanapproachwithpropertyofreliable．highe伍cien

4、tandeasyrealization．Projectgroupproposedastructuredlight3DdetectingapproachbasedonsymbolicM·arrays，whichaimedatdynamic3Dscenereconstructionusingbinarycodedlightpattern．Asoneofkernelsofstructuredlighttechnique。thecodificationofstructuredlightpattemandthedecodi

5、ficationofsymbolimageaffectthedetectionSuccessratio．resolution,precisionandrobustnessdirectly．Meanwhile，thedecodificationofsymbolimagereliesontargetsurfacereflectivity,surfacegradiem，andhardwarenoises．Especially,duetothepropertiesoflargesize，elements，denseofk

6、eypoints，largeinformationofthestructuredlightpatternproposedbyresearchgroup，itistheoneofkeyproblemstodecodethelightpatternimagequickly,retrievesymbolcodewordse伍cientlyandlocatekeypointsprecisely．Aimingatthecodificationsymbolsandcodificationpattemwhicharenewly

7、proposedbyresearchgroup，anoveldecodingapproachforsymbolicM—arraysstructuredlightimageisproposedinthispaper．Basedonextractingthecrosspointsneteffectively,theproposedapproachimplementedsymbolsrecognition,windowmatching，codewordretrieving，keypointslocationandtar

8、getsurfacereconstruction．ThefilterapproachofreconstructedobjectsurfaceisalsodiscussedinthisP印eLThemaincontextofthisP印ercanbesummarizedasfollows：(1)StudyonSymbolimagepreprocessing．Severalt