Abstract:

A method was established for the automatic discrimination of three varieties of fishmeal by means of near infrared
spectroscopy (NIRS). Through analysis of the spectral differences of fishmeal samples, a discrimination model for different
types of fishmeal was developed by using t principal component analysis (PCA). The spectra were scanned from 1 100 nm
to 2 498 nm. The 1 minus variance ratio (1-VR) was 0.913 5 and the standard error of cross validation (SECV) was 0.133 8.
The accuracy rates of discriminate for calibration and external validation were 84.6% and 100%, respectively. The
results of the study indicate that NIRS combined with chemometrics is rapid, nondestructive, reliable and suitable for the
discrimination of three varieties of fishmeal.

Key words: near-infrared spectroscopy, fishmeal, discrimination, varieties, principal component analysis