织物表面折皱的小波分析与自组织神经网络等级评定

《织物表面折皱的小波分析与自组织神经网络等级评定》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、第10卷第4期中国图象图形学报Vol.10,No.42005年4月JournalofImageandGraphicsApr.,2005织物表面折皱的小波分析与自组织神经网络等级评定杨晓波黄秀宝(浙江财经学院信息学院,杭州310012)(东华大学纺织学院,上海200051)摘要为了提取较为精细的图像信息,引入了多尺度2维小波分析织物的表面折皱。织物图像首先经过高斯滤波,再利用小波变换分解并从中提取高频信息,然后结合4种表征织物折皱的特征参数,计算不同折皱等级模板的特征值,通过分析特征值与折皱等级的相关系数,

2、表明这4种特征参数可以作为模式识别的输入量;最后采用Kohonen自组织神经网络客观评定织物的折皱等级,自组织神经网络将不同等级的织物折皱模板进行分类,并以此为依据,确定26种不同织物类型的折皱等级。为了定量描述评定结果,通过计算客观评定与主观评定结果的相关系数,验证该方法的可行性。关键词小波分析特征提取折皱等级评定中图法分类号:TN701文献标识码:A文章编号:100628961(2005)0420473206WaveletAnalysisofFabricSurfaceWrinkleandSelf2or

3、ganizedNeuralNetworkGradeAssessmentYANGXiao2bo(DepartmentofInformation,ZhejiangUniversityofFinance&Economics,Hangzhou310012)HUANGXiu2bao(CollegeofTextiles,DonghuaUniversity,Shanghai200051)AbstractInthispaper,Multi2Scaletwo2dimensionalwavelettransformisimp

4、ortedtoanalyszefabricsurfacewrinkleinordertoacquirethefinerimageinformation.Firstly,fabricimagecanbefilteredthroughGaussianfilter,anddecomposedbywavelettransform;meanwhile,highfrequencyinformationisextracted.Secondly,fourkindsofwrinklefeatureparameterarea

5、ppliedtocalculatethefabricwrinkledegreewithdifferentwrinklereplica,whicharehorizontalvariance,verticalvariance,horizontaloffsetandverticaloffsetseparately.Throughanalyzingthecorrelationcoefficientbetweenfeatureparameterandwrinklegrade,whichindicatesthefou

6、rkindsofwrinklefeatureparametercanbetakenastheinputvalueforpatternrecognition.Finally,Kohonenself2organizedneuralnetworkisalsousedtoevaluatefabricwrinklegradeobjectively.ThewrinklefeatureparametersareinputtotheKohonenself2organizedneuralnetwork,throughtra

7、iningandstudyingprocess,theoutputvaluecanbeobtained,differentwrinklegradeoffabricreplicawillbeclassifiedbyapplyingself2organizedneuralnetwork,andwrinklegradeof26differenttypefabricscanbeevaluatedaccordingtothisresult.Fordescribingtheassessmentresultwithqu

8、antify,thecorrelationcoefficientiscalculatedbetweenobjectiveassessmentandsubjectiveassessmentinordertovalidatethefeasibilityofthismethod.Keywordswaveletanalysis,featureextraction,wrinklegradeassessment基金项目:浙江省教育厅科研项