Vibrational spectroscopy in the near infrared (NIR) has many applications in industry, particularly in quality control, discrimination and determination of the chemical composition of different foods because of the speed and ease of application in routine analysis when compared to traditional physical and chemical methods. This work aims to use this technique to discriminate five fish species from the Amazonian coast (Genyatremus luteus , Lutjanus purpureus , Macrodon ancylodon , Cynoscion acoupa, Micropogonias furnieri) and predict major components (water content, lipids and proteins). A portable low-resolution spectrometer (1600-2400 nm) was used for obtaining spectral measures of fish samples (intact, crushed and lyophilized muscle tissue). The spectra were preprocessed by means of derivatives using the Savitzky-Golay algorithm for smoothing of spectral noises, when constructing both prediction and discrimination models. The PCA and SIMCA models were applied for species discrimination, while PCR and PLS were used in the prediction of the major components. Using PCA, it was observed the formation of well-defined groups of the species G. luteus and L. purpureus in intact and lyophilized samples. With SIMCA, it was observed the formation of groups of the five species, which were confirmed by the distance between groups in the range from 4.16 to 13.31 and 2.90 to 57.05 for intact and lyophilized samples, respectively. The PLS regression model showed a small positive variation values for r2 and R2 , and the lyophilized samples showed better results for the three studied parameters: water content ( r2: 0.58 , RMSEPcv: 1.11 and RDP: 1.40 ), lipids (r2 : 0.96 , RMSEPcv : RDP 1.09 and 4.82 ), and crude protein (r2: 0.86 , RMSEPcv : RDP and 1.96 : 2.60). A good correlation between proximate composition data and those predicted by PLS regression was obtained.
|
