1、中國農(nóng)業(yè)大學碩士學位論文近紅外光譜技術(shù)評定綿羊采食性和日糧營養(yǎng)成分的研究姓名：舒彬申請學位級別：碩士專業(yè)：草業(yè)科學指導教師：張英俊20070601AbstractNearinfraredreflectancespectroscopy州ms)isaneffectiveandnondestruetiveanalyticalmethodfororganiccompoundSincetheintroductionofNIRSinthe1960s

2、。thenumberofNIRSalded刪yticalapplicationshasgreatly“pandedTheNIRStechnologyhasbcellwidelyusedtoassessnutritionvaluesforherbivoresinrecentyearsHowevertherehasbeenlittlepublishedexperimentofusingNIPStechniquetopredicttheint

3、akeanddietcompositionofanimalsgrazinginthetypicalsteppegrasslandThestudyWasconductedtodetermineiffecalNIPSisabletopredictdietcompositionanddietqualityofforagefedsheepandevaluatetheperformarmoffecalNIPSequations1Theextern

4、alvalidationresultsshowedthatthec∞銜cientofcorrelationofprediction(0forDMIandtheproportionofLo“l(fā)口chinensisElymussibiricusPuccinelliadistansandChenopodiumalbumindietswe”respectivelyO6896。089020，7344，O8624andO8295andtheroot

5、meansquareerrorofprediction(RMSEP)forDM!andbotanicalcompositionsindietswele10709DM)，1020(％DM)1110(％DM)，844(％DM)and376(％DM)，respectivelyTherewaslittledifferencebetweentheaveragevaluesofDMIandbotanicalcompositionsobtainedb

6、yusingfecalNIRSpredictionandbyusinglaboratoryanalysisThefecalNIRSgavea訓sfktofyaccuracyintheanalysisoftboproportionsof]eyrau$chinensisPuccinelliadistansandChenopodiumalbuminthisstudyThepoorperformanceofpredictionforDMIsho

7、wedth越thecalibrationmodelbasedonsmallcalibrationsampiesetfn=35)wouldprovidealoweraccuracyinthepredictionforunknownsampleThepoorqualityofcalibrationforChenopodiumalbumWasprobablyduetotheamountofthiscomponentlessthan10％int

8、hedietThepoorqualityofpredictivemodelforElymussibiricusshowedthesp％Iralinformationmeasuredfromtheplantofthesamespeciesor0chemicalcharacteristicsmaybecoincidencethehomogeneityofthecompositionsanalyzedwouldaffectspet：trali

9、nformationextractionespeciallyinsmallsampleset2Theextemalvalidationresultsshowedthatthecoefficientofcorrelationofprediction(r)forC只ADFandNDFindietswererespectively0892700894and07“8andtherootmeansquareerrorofprediction(RM

10、SEP)forDMIandbotanicalcompositionsindietsWgre087(％DM)，350(％DM)and233(％DM)respectivelyThefecalNIRSmodelsdevelopedinthisstudygaveasatisfactoryaccuracyintheanalysisofCPandNDFThepoorperformanceofpredictionforADFshowedthatthe

11、NIRSequationscannotbeextrapolatedbeyondtheconditionsrepresentedincalibrationsamplesandlittlevarietyofsampleswouldreducethequalityofNlRSpredictiveequationTheresultsconfirmedthatthepotentialoffecalNIRSasapromisingtechnique

12、tomeasuredietarycompositionanddietqualityofforagefedsheepHoweveradditionalworkinthisargamayberequiredtodetermineiffecalNIRSpredictionsofnutritionvaluesCallattainhigherlevelsofaccuracyandprecisionKeywords：Nearinfraredrefl