Freshness evaluation model of Pseudosciaena crocea based on near-infrared spectra

Objective To investigate a method for the quantitatively freshness evaluation of Pseudosciaena crocea.Methods Near-infrared spectra of the whole back of fish was adopted and preprocessed.Quantitative models of total volatile basic nitrogen(TVB-N) content and aerobic plate count were built with the processed spectra,respectively.The partial least squares(PLS),interval PLS(iPLS),backward interval partial least squares(biPLS) and synergy interval partial least squares(siPLS) algorithms were used for modeling.Results biPLS model had the highest accuracy and predicted the best performance.The optimal biPLS model of TVB-N was achieved with correlation coefficient(R c =0.8371) in calibration set and correlation coefficient(R p =0.7652) in prediction set.The optimal biPLS model of aerobic plate count was achieved with correlation coefficient(R c =0.878) in calibration set and correlation coefficient(R p =0.7009) in prediction set.Conclusion There is a high correlation between near-infrared spectra and TVB-N or aerobic plate count.Near-infrared spectroscopy with biPLS can be successfully applied as an accurate and non-destructive method for the determination of freshness of Pseudosciaena crocea.