欢迎来到天天文库
浏览记录
ID:43712749
大小:1.12 MB
页数:35页
时间:2019-10-13
《红外焊缝图像识别算法研究》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、上海交通大学硕士学位论文红外焊缝图像识别算法研究姓名:胡胜华申请学位级别:硕士专业:精密仪器及机械指导教师:丁国清20060101和不规则性判断三种识别算法对红外图像进行检测识别针对焊缝特征的多样性和噪声的不确定性充分发挥各种算法的优势提高焊缝识别的准确率其中不规则性判断是结合红外焊缝图像识别的具体实践模拟人的眼睛识别红外焊缝图像的过程而提出的识别算法其依据为焊缝形成时的瞬时高温打乱了丁件原来比较规则的纹理形成了相当明显的不规则性从而造成已焊图像和未焊图像在不规则性上的差别所有红外焊缝图像识别处理过程都在WINDOWS操作系统上采用VC6.0编程实现并进行了大量的试验试验
2、结果证明此红外焊缝图像识别算法有相当高的准确率其中7条焊缝的判断准确率基本达到Tioo关键词灰度共生矩阵KL变换不规则性判断第II页WELDINGLINEIMAGEDETECTALGORITHMRESEARCHABSTRACTAlongwiththefastdevelopingofmodernindustryasmetallurgy,machinerypetrifaction,theatom,spaceflightandsoon,theweldingbecomemoreandmoreimportant.Ithasbecomethethirdcriticalkindoffa
3、bricationindustry,thatisonlylessimportantthanassemblingandmachinerymachining,sotheassuranceoftheweldingqualityisofbigmagnitude.Thequalityoftheweldinglineincludesshape,positionandfault.Especiallyintheautoindustry,itisimportanttodetectandestimatetheweldingline.Howtodothisfastandexactlyisthe
4、criticalproblemtosolve.ThecontentofthisthesisisoneofthekeytechnologiesintheWeldingLineDetectSystem,whichisaprojectofShanghaiAutoFund.Wedevelopedthissystemtodetectthequalityofthe21weldinglinesinBuickautomobileauxiliaryframe.Becausethepositionandtheshapeofthe21weldinglinesisdifferent,number
5、sofCCDcamerasareusedtogetimages.AllofthesecamerascommunicatewithcomputerviaUSBinterface.Then,weprocessthecapturedimagesandestimatethequalityofWeldinglineconsequently.Thisthesisfocusesontheresearchofinfraredweldinglineimagedetectmethodandisorganizedbyinfraredweldingimageprocessflow.Atfirst
6、,weanalyzethecharacteristicsofweldinglineimage.Thenwepre-processtheweldinglineimagewiththemethodofimageenhancementsuchasbrightnessadjustmentandfiltering・Intheinfraredweldinglinedetectmodule,weusedGreyLevelcooccurencematrix,KLtransform,andtheirregularfeaturesmethodtodetecttheweldingline.Wi
7、thinthesemethods,theirregularfeaturesmethodisanewmethodsimulatinghuman'seyestoidentifyweldingline,which第hi页isbasedonafactthatthehightemperaturewilldestroysteel,stexture.AlltheinfraredweldinglineimagedetectstepsareaccomplishedwithVC6.0onWINDOWSOS.Theexperimentalresul
此文档下载收益归作者所有