Prediction of meat and bone meal content in feed by near infrared reflectance spectroscopy

Twenty-eight feed samples for chicken, 31 feed samples for swine, 25 feed samples for cattle and 7 meat and bone meal(MBM) samples were collected, and MBM was deliberately adulterated with feed at 0.5%～6.0% by weight, and analysis samples were prepared. The initial spectrum was pretreated by scatter correcting, smoothing, first derivative and second derivative, respectively. The NIRS calibration models of the prediction meat and bone meal content in feed for chicken, swine and cattle were developed using the partial least squares(PLS) regression technique. The calibration model was proved in its precinct by validation set samples. The coefficients of determination in calibration sets (R2 ) are 0.9694, 0.9846 and 0.9788; the standard errors (RMSEC) are 0.279, 0.252 and 0.287; the relative percent differences (RPDs) are 5.663, 6.865 and 5.889 for chicken, swine and cattle feed, respectively. The results show that the NIRS can accurately and quantitatively measure the MBM content in feed.