Abstract
The results obtained in our experiment are two-fold : Firstly, a convenient model of a near-infrared spectrophotometer, which has proved to be useful in the determination of mellowness in fruit, was successfully applied to the measurement of fat content in horse mackerel and dried split horse mackerel. This measurement may be performed instantly, with the specimen remaining intact. To obtain this result, we used multiple linear regression (MLR) based on second derivative values of spectra measured at the back abdomen side part of the fish body and the fat content of the whole body. This resulted in a good calibration equation, with D2 log(1/R), at the fat band of 924nm, as the first variable. Secondly, we succeeded in developing a global calibration equation that can be used for both frozen and thawed samples of horse mackerel, eliminating the need to measure the temperature of each specimen. The global calibration equation has the function of temperature compensation, with high accuracy : 0.88 decision coefficient (R2) and 1.6% standard error of prediction (SEP) for whole fish (actual fat range 1-24%) and dried split fish (actual fat range 1-19%).
