Using NIR Spectroscopy to Discriminate Surimi Made from Fishes Stored in Different Condition

Abstract. In this experiment, we collected the near-infrared (NIR) diffuse reflectance spectra of the frozen surimi made from grass carps stored in different conditions. PCA (Principal component analysis), DA (discriminant analysis) and PLSDA (partial least-squares discriminant analysis) were applied to discriminate different surimi samples. The results showed that NIR spectroscopy combining chemometrics methods could classify different surimi samples. The correct classifications were 97.67% and 98.84% for DA model with raw spectra and PLSDA model with spectra experienced MSC and the second derivative pre-treatment. For PLSDA model the best satisfactory calibration and prediction abilities, with rc = 0.9889, RMSEC (root mean square error of calibration) = 0.024, rv = 0.9630, RMSEV (root mean square error of validation) = 0.044.