LAUSR

Hyperspectral imaging (HSI) was investigated to detect and visualize beef adulteration with duck meat. Minced beef was adulterated with duck meat in range of 0–100% (w/w) at 10% increments. Visible and near-infrared (VNIR) hyperspectral images were acquired, and extracted spectra were analyzed with preprocessing methods. Partial least squares regression (PLSR) and principal component regression (PCR) modeling methods were individually utilized, and the PLSR model based on full raw spectra performed best. To make the preferred model practical, optimal wavelengths were individually selected using two-dimensional correlation spectroscopy (2D-COS) and PC loadings. The model based on competent wavelengths selected from PC loadings resulted in good performance of Rp2 = 0.96, RMSEP = 6.58%, and RPD = 4.86 with limit of detection (LOD) of 7.59%. Finally, spatially distributed visualization was achieved using the simplified model, and adulteration levels were readily discernible. Known distributed samples were also successfully visualized, and the validity of visualization was thus proved. Results demonstrated the potential of the HSI to detect minced beef adulteration with duck meat.