Qualitative Identification of Fish Meal and Meat Bone Meal via Fluorescence Spectral Imaging

In order to improve feed safety, we investigated the feasibility of using fluorescence spectral imaging (FSI) technique to qualitatively detect meat and bone meal (MBM) mixed with fish meal (FM). Using the FSI system, the spectra of 90 FM and MBM samples from different regions were collected with denoising and binary processing to obtain images. The spectral data curves over the wavelength range of 400-680 nm were used for identification and were pre-treated with various methods. Next, the curves were plotted based on stereo spectrograms; then, partial least squares discriminant analysis (PLS-DA) was applied to determine the FM and MBM samples. One FM sample and one MBM sample were randomly selected and mixed together (v/v, 1:1), and the mixture's pseudo-color (PC) image was generated by mapping the best grayscale values using a spectrogram method. The resulting PC image was then used for spatial identification of the mixture. The PLS-DA results indicated that the identification effect was best after the multiplicative scatter correction where the sensitivity, specificity, and accuracy all equaled to 1. The PC image demonstrate that the FM (green) and MBM (red) accounted for 52.6 and 47.4 % of the total volume, respectively, corresponding with the actual 1:1 ratio. Additionally, the experimental test of three low concentration-mixed powder samples (c[MBM] = 1.5 %, c[MBM] = 3 %, and c[MBM] = 5 %) determined that the lower limit of application (LLA) of the proposed method is c[MBM] = 1.5 %. Theoretical limit of detection (LOD) of the method was calculated about c[MBM] = 0.75 % through linear regression. These experiments show that the spectral and spatial information provided by FSI can rapidly and nondestructively determine whether FM is mixed with MBM.