Study On The Determination Of Cottonseed By Near-Infrared Spectroscopy

The research studied the methods and ways of the application of near infraredspectral analysis technology,which determined the water content, protein content,NDF content, ADF content, crude fat in cottonseed rapidly and accurately:(1)exploredthe influence spectrum pretreatment method (mathematical treatment, scatteringcorrection) to the calibration model, compared the application of the three regressiontechnology (modified partial least square method, and partial least square method,principal component regression method) in the modeling;(2)researched the feasibilityof the Foss XDS near infrared analyzer was applied to determine the cottonseedquality: moisture content, protein content, NDF content, ADF content and crude fatcontent, and set up the near infrared calibration models for these ingredientsrespectively;(3)choosed a group of samples which had nothing to do with thecalibrations, validated the optimal model, through the comparison of the differencebetween predicted value and chemical determination value could explain the accuracyof the optimal model. The results showed that:(1) For Foss XDS near infrared analyzer, the modified partial least squaresmethod is the most suitable regression technique for the establishment of thecalibration model water content, protein content, NDF content, ADF content, crude fatcontent.(2) With the calibration models of moisture content, protein content, NDF content,ADF content and crude fat content by Foss XDS the standard error of cross-validation(SECV)were0.0924,0.2197,0.2457,0.2481,0.3657; the modelsdetermination coefficient of correlation(RSQ)were0.998,0.991,0.930,0.972,0.992;the square error of prediction(SEP)were0.108,0.231,0.201,0.091,0.445; T test valuebetween the chemical standard methods and NIRS method were0.00008,0.00016,0.00018,0,0.0985(P<0.05), this showed that the two methods had not statisticdifference. This NIRS method can be applied to the collection quality of cottonseed.