Abstract:

Total volatile basic nitrogen (TVB-N) content of Yao meat stored at 4 ℃ after the package was opened was
measured every 24 h for 8 days, and reflectance spectra were collected from the hyperspectral scattering images (HIS) in the
range of 430 to 960 nm. After pre-processed by standard normal variate transformation (SNV) method, prediction models
for TVB-N content in Yao meat were established by partial least squares (PLS) method, interval partial least squares (iPLS)
method, backward interval partial least squares (biPLS) method and synergy siPLS method separately. Experimental results
showed that the siPLS model could predict Yao meat TVB-N with correlation coefficient (Rp) of 0.854 8 and root mean
squared error of prediction (RMSEP) of 2.47, which was the best of the four models, and 430–461, 555–586 and 929–960
nm were the selected wavelength ranges. The overall prediction accuracy of the siPLS model for Yao meat freshness could
reach up to 87.5%. The present study shows that HIS technique is feasible for fast and non-destructive detection of Yao meat
freshness attributes.

Key words: hyperspectral imaging technique, Yao meat, freshness, partial least squares (PLS)