Identification of main meat species based on spectral characteristics

Abstract. The identification of meat species and the determination of meat authenticity have drawn an increasing amount of attention for researchers and consumers. Beyond doubt, the rapid discrimination of meat plays an important role in the scope of food safety and quality control. In this study we investigated the identification of three different meat species including pork, beef and lamb by applying hyperspectral imaging technology. This attributes to that various kinds of meat have their own spectral characteristics. First of all, hyperspectral images of different kinds of meat samples were acquired. After that, the calibration of acquired images and ROI selection were carried out, and then, four different spectra pre-processing methods including first derivative, multiplicative scatter correction (MSC), standard normalized variate (SNV) and Savitzky-Golay smoothing were used. At last, in order to develop discrimination models, the Fisher linear discriminant method was applied to identify the meat species with the characteristic wavelengths, selected by stepwise regression (SR). The calibration set included 46 samples (pork=15, beef=15 and lamb=16), and another15 samples (pork=5, beef=5 and lamb=5) were used to validate the identification effect. The 100% identification accuracy was obtained using the optimum wavelengths selected by SR for the spectra after SNV pre-processing. The results demonstrated that the rapid and non-destructive hyperspectral imaging technology based on the spectral characteristic could serve as an effective identification method for meat species.