Combining Vis-NIR and NIR hyperspectral imaging techniques with a data fusion strategy for the rapid qualitative evaluation of multiple qualities in chicken

Total volatile basic nitrogen (TVB-N) and total viable count (TVC) are two important indicators for evaluating and controlling the quality and safety of chicken fillets. The aim of this study is to evaluate two hyperspectral imaging techniques (visible near-infrared (Vis-NIR) and NIR) in combination with low-level and intermediate-level fusion strategies (LLF and ILF) for the prediction of multiple quality indicators of chicken fillets stored at 4 °C. Quantitative predictions using partial least squares regression (PLSR) were established after preprocessing and feature wavelength selection. The results showed that the data fusion strategy exhibited better performance in both indicators compared to individual data. The LLF strategy showed optimal performance in predicting the TVC content with an RP2 of 0.9275 and an RMSEP of 0.1889, while the ILF strategy was best in predicting the TVB-N content with an RP2 of 0.8652 and an RMSEP of 2.6094. Moreover, the optimal models based on selected bands were used to achieve visual maps of the TVB-N and TVC contents. Although validation with an independent batch of samples was not used, it was a feasibility and valuable study. The experiment results demonstrated that the fusion of two types of hyperspectral data can be successfully used to evaluation of chicken multiple qualities, and provided a potential method for monitoring and detection of meat qualities. [Display omitted] •Two types of HSI techniques were used and combined to predict TVB-N and TVC.•Preprocessing and feature extraction algorithms were employed to seek the optimal combination.•The superiority of low- and intermediate-level fusion strategies was demonstrated.•Distribution maps were visualized to observe the changes of chicken qualities.