The calibration model of the pork quality based on hyperspectral data was easily affected by pork varieties and had poor applicability. So different algorithms were compared and a calibration transfer algorithm based on spectral value correction was proposed to improve the model applicability between different varieties. The PLS model for Shanhei pig was established, then was used to predict Linghao pig samples. Prediction accuracy was very poor with only Rp of 0.415, and RMSEP of 0.1804. To improve model applicability, S/B algorithm, model updating and transfer algorithm based on spectral value correction were respectively adopted and compared. For S/B algorithm, Rp of the model prediction for Linghao pig samples was still 0.415, and RMSEP decreased from 0.1804 to 0.1343, only down by 25.54%. For the model updating method, when 14 Linahao pig samples were added to the calibration dataset, the model prediction performance for Linghao pig samples achieved optimal. Rp increased to 0.797, improved by 92.05%, and RMSEP reduced to 0.1121, dropped by 37.86%. For transfer algorithm based on spectral value correction which combined spectral value physical value coexist distance with DS algorithm, Rp of the model prediction for Linghao pig samples increased to 0.837, rose by 101.69%, and RMSEP reduced to 0.0856, fell by 52.55%. The results showed that the transfer algorithm based on spectral value correction could eliminate the difference of spectral value of different varieties efficiently and improve the model applicability. It gave the best transfer and correction result than the two other algorithms.