Abstract:

Protein, fat and moisture in 107 grass carp samples were analyzed by chemometric methods. Meanwhile, near
infrared spectra (NIRS) of these samples were investigated in order to obtain quantitative analysis methods for grass carp
nutrients. The protein content in grass carp ranged from 18.00% to 26.00%, fat content from 2.08% to 4.36%, and moisture
content from 70.80% to 81.20%. The results showed that the nutrient values met the modeling requirements. Various
spectral preprocessing methods were tested, and the best preprocessing method was multiplicative scatter correction (MSC)
together with Savitzky-Golay derivatives for grass carp moisture spectra, and standardization for protein and fat spectra. The
near-infrared quantitative analysis models were obtained for fish protein, fat and moisture contents by partial least square
regression. The correlation coefficients of the models were 0.9806, 0.9968 and 0.9372 for protein, fat and moisture contents,
respectively. These results indicate the acceptable fitting accuracy and prediction capability of the models for the analysis of
protein, fat and moisture contents in grass carp by NIRS.

Key words: near infrared spectroscopy, protein, fat, moisture, quantitative mode