雞蛋外觀品質(zhì)識別方法及試驗裝置研究.pdf

1、內(nèi)蒙古農(nóng)業(yè)大學(xué)碩士學(xué)位論文雞蛋外觀品質(zhì)識別方法及試驗裝置研究姓名：張曉芳申請學(xué)位級別：碩士專業(yè)：機械設(shè)計及理論指導(dǎo)教師：郁志宏20060501ResearchonRecognitionMethodsofEggExteriorQualityandExperimentDeviceAbstractTheexperimentdeviceisdesignedfordetectingeggexteriorqualityappropriateligh

2、tsourceandbackgroundareselectedThecapturedimagesarepre—processedwiththemethodsofdecliningnoise、imageenhancement、binarization、edgedetection，etcIterativemethodisadoptedtoobtainthebestsegmentationthresholdwhileimagebinariza

3、tion，andoptimumOstualgorithmisputforwardforboundaryextractionFordirty、crackimagessegmentation；firstlyOstualgorithmiscombinedwithsobeloperatorasaresultinternalinformationissupplementedonthebasisofguaranteeingclearimageedg

4、e，andthenthresholdrecognitionmethodisusedforgettingridofinterferentialvalueandlines，theaccuracyofextractingthefaultfeatureiselevated，atlasteggsaredetectedaccordingtothenumberofwhitepixelsoccupiedbyfaultsExtractingfeature

5、parametersofapparentquality，areaofeggimageisusedtoevaluatingeggweight，thecorrelationcoefficientisO9922Eggformindexisselectedasshapemeasurement，whencalculatingeggformindex，modifiedminimumrectangleisproposedforcalculatinge

6、gglongandshortaxes，theaxesandactualmeasurementvaluesarecorrelatewith09733、O9797BasedonanalysisofcomparingthestandardBPandthreekindsofmodifiedBPalgorithms，thestrategyisadvocatedthatfirstlyneuralnetworkweightsareoptimizedt

7、hroughGAalgorithm，andthentrainingBPnetworkbyL—Malgorithm，networktrainingconditionsareimprovedvalidlyNetworksimulationiscarriedonwithexaminationsamples，theclassificationaccuracyofGA—LMnetworkis938％foreggweightand100％foreg

8、gshapeKeywords：Machinevision：EggqualityiArtificialneuralnetwork；Geneticalgorithm；Directedby：APYUZhihongAppIicantforMasterdegree：ZHANGXiaofangfMechaflicaIDesignandTheory)(CollegeofMachineryandElectricalEngineeringInnerMon