[Research on predicting the nutrition value of pelletized alfalfa by near infrared reflectance spectroscopy].

The present research aimed to predict the qualities of pelletized alfalfa by near infrared reflectance spectroscopy. Sixty pelletized alfalfa samples were collected, including 22 whole plant alfalfa samples, 19 stem samples and 19 leaf samples. They were divided into a calibration sample set (45 samples) and a validation sample set (15 samples). The Fourier transform-near infrared reflectance spectroscopy (FT-NIRS) and the partial least square (PLS) were used to calibrate models of the pelletized alfalfa nutrition value, involving crude protein (CP), neutral detergent fiber (NDF) and acid detergent fiber (ADF) contents. All models had great calibration performances. The correlation coefficients of cross-validation (R(CV)) were between 0.96410 and 0.96887, and the root mean square errors of cross-validation (RMSECV) were between 0.80% and 2.59%. Fifteen validation samples were used to predict the performances of these models, all the correlation coefficients of NIRS value and chemical value (r) were between 0.9669 and 0.9743, and the root mean square errors of prediction (RMSEP) were between 0.85% and 2.07%. The RPD values of cross-validation and prediction were all above 3. The results showed that pelletized alfalfa's CP, NDF, ADF contents were exactly predicted by near infrared reflectance spectroscopy.