基于遺傳算法和支持向量機(jī)的混紡紗線橫截面圖像分割方法的研究.pdf

1、南京理工大學(xué)碩士學(xué)位論文基于遺傳算法和支持向量機(jī)的混紡紗線橫截面圖像分割方法的研究姓名：林森申請學(xué)位級別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：周凌柯20100620碩士論文AbstractTextileisoneofthemostimportanttraditionalindustriesinChinaofwhichtextiletestingisthecriticalpartMostofthetraditionaltextilete

2、stingreliesonmanualoperationandsubjectiveassessmentHowtoimprovetheobjectiveassessmentandthedegreeofautomationintestingisthekeypointinachievingsustainabledevelopmentinthisfieldTherapiddevelopmentofthecomputergraphicproces

3、singandthemachinelearningtechnologyispouringinfreshvitalitytomanyfieldsHowtocombinethecomputergraphicprocessingtechnologywiththemachinelearningtheorytogetabettertestingresultisthemostimportanttechnicaldifficultyThisartic

4、lemainlycentersonthemethodtotrainthesupportvectormachine(SVM)toseparateyamsintotwocategoriesofwoolandviscosebycatchingthecrosssectiongraphicofwool／viscoseblendedyamsThemainresearchisasfollows：(1)Acommonlyusedgraphicprepr

5、ocessingflowwasproposedbasedonthesegmentobjectsstatedinthisarticle(2)Anextractionwasputforward，judgingbytheircharacteristicsofwoolandviscoseintheblendedyamInordertoobtainthesamplecollectionusedforthetrainingofSVM，acompar

6、isonwassetupbetweentheircharacteristicsandthefollowingaspectssuchasregularcircle，ellipse，rectangleandcircumscribingpolygons(3)Abasicframeworkwasdiscussedaboutthegeneticalgorithm(GA)andtheSVMWiththeaidoftheframework，these

7、lectionoperatorofGAwasimprovedandthetypeofkernelfunctionoftheSVMwasselectedaswellThisimprovementandselectionwasprovedreasonablebyanapplicableexample(4)Aseparationmethodwasanalyzedthroughdemonstrationonthebasisoftheimageo

8、fGAandSVMGroundedontheapplicationofseparationaccordingtothecrosssectiongraphicoftheblendedyam(woolandviscose)，itprovedthatthisseparationmethodisfeasible(5)Acomparativeanalysisofparameteroptimizationbetweenanonheuristic鰣d

9、searchandaheuristicgeneticalgorithmWasdealtwithItmainlycomparativelyanalyzedthespeedofobtainingthecategorizermodelbetweenthemTheresultisthecombinationofthegeneticalgorithmandtheSVMmethodisappreciatedIntheend，thestructura