Prediction of TVB-N content in beef with packaging films using visible-near infrared hyperspectral imaging

A practical prediction method for the Total volatile basic nitrogen (TVB-N) content evaluation of beef covered by the packaging films was proposed by using visible-near infrared hyperspectral imaging (HSI). Hyperspectral image under three conditions: beef without films, beef covered by appressed films, and beef covered by non -touch films was obtained. The results indicated that the influence on prediction accuracy of TVB-N content with the presence of films could be properly eliminated by spectral preprocessing methods and modeling methods. Moreover, the successive projections algorithm (SPA) combined with least squares support vector machines (SPA-LSSVM) was the best predictive model for beef without film, beef covered by non-touch poly-propylene and polyethylene, and determination coefficient in prediction (R2P ) were 0.9167, 0.9157, and 0.9423 respectively, and root mean square errors in prediction (RMSEP) was 1.7808, 1.8758, and 1.4304 respectively. The competitive adaptive reweighted sampling (CARS) combined with LSSVM (CARS-LSSVM) was more effective for beef covered by appressed films of polypropylene and polyethylene, and R2P were 0.9046 and 0.9533 respectively, and RMSEP was 2.0444 and 1.2858 respectively. This study demonstrated that it is feasible to predict the TVB-N content of packaged beef using visible-near infrared hyperspectral technology.

Automated summary

A practical prediction method for evaluating the Total volatile basic nitrogen (TVB-N) content of beef covered by packaging films was proposed using visible-near infrared hyperspectral imaging (HSI). The study demonstrated that the influence of films on the prediction accuracy of TVB-N content can be eliminated by spectral preprocessing and modeling methods. It proved the feasibility of using visible-near infrared hyperspectral technology to predict TVB-N content in packaged beef.