抄録
The near-infrared (NIR) spectrum of fish meal was determined on a scanning spectrophotometer. The multiple linear regression (MLR) and partial least squares (PLS) regression were used for the calibration of moisture, protein and fat contents in fish meal. Both the MLR and PLS gave high correlation coefficient and low standard error of calibration. Correlation coefficients of calibration were more than 0.90 for each component. The prediction results also show high correlation coefficients (more than 0.90). The relative errors for prediction of each component were less than 9%, and for most samples were less than 6%. Therefore, NIR spectroscopy could be a useful technique for simultaneously determining moisture, protein and fat contents in fish meal. The time required for analysis of one sample is less than 3 min. It is a rapid and nondestructive method without environmental pollution. Thus, it will result in considerable time-saving, and it can be applied to on-line analysis at fish meal processing sites.
