| Taking mirror carp as the research object, the near infrared spectroscopy (NIRS) and stoichiometry
were used to collect and determine the related indexes. The partial least squares (PLS), partial least squares (PLS) and
BP artificial neural network were used to optimize the model. The optimum modeling method, pretreatment method
and wave band were determined under 21 pretreatments by scanning the spectrum of fish and pH value, TVB-N (volatile
base nitrogen) value and TBA (thiobarbituric acid) value. The model optimization shows that pH, TVB-N and TBA
are the best models in the partial least squares method. The optimal preprocessing methods are baseline correction
and standard normal variable transformation, net analysis signal, Savitzky-Golay derivative and baseline correction.
The optimal bands respectively are 1000-1300 and 1700-1799 nm, 1000-1200 and 1300-1650 nm, 1000-1799 nm.
The RC values of pH, TVB-N and TBA respectively were 0.9906, 0.99865 and 0.99971, respectively. The Rp values
respectively were 0.6436, 0.021357 and 0.7723. The establishment of a quantitative detection and prediction model
for freshness of mirror carp by near infrared spectroscopy was achieved. |
