Enhancement of species-specific analysis for meat and bone meal by matrix fragments-related spectral fusion

Abstract

In this investigation, a pioneering approach involving the fusion of matrix fragments-related spectral data was proposed to improve the underperformance observed in raw meat and bone meal (MBM) when employed for species discrimination analysis. Initially, the MBM matrix was characterized as a binary mixture comprising bone fragment (BF) and meat fragment (MF). Subsequently, the disparities in near infrared (NIR), mid infrared (MIR), and Raman spectra between BF and MF samples were individually identified and elucidated. Following, the spectral fusion data related to matrix fragments were synthesized and subjected to analysis using principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) for species-specific evaluation. The suggested data fusion strategy was authenticated by its capacity to facilitate improved differentiation within the principal component space, along with reduced classification errors in PLS-DA. Further, complementarity of matrix fragments-related spectral variables for MBM species discrimination analysis was explicitly scrutinized and contributions to MBM derived from four species were meticulously traced. Additionally, the proposed analytical strategy for MBM could serve as a reference for the spectral characterization of other agricultural materials with complex matrices.