Abstract: A PLS model was constructed to predict the 2-thiobarbituric acid (TBA) value in chilled chicken based on optimal spectral information selected by PLS-β method, and to evaluate the oxidation degree of chilled chicken in a rapid and non-destructive way. The hyperspectral images of chilled chicken samples in the range of 900~1700 nm were acquired and the spectral reflectance information within the region of interest of the images were extracted and averaged. Eight methods, including MAS, SGS, MFS, GFS, normalization, MSC, BC and SNV were applied for spectral pretreatment. And then, partial least squares (PLS) models were built to predict reference TBA value in chicken. The PLS-β method, stepwise and successive projections algorithm (SPA) were respectively adopted to select optimal wavelengths to simplify the PLS model. The results indicated that the GFS-OW-PLS model established based on 31 optimal wavelengths selected from the GFS spectra by PLS-β method had best performance in predicting TBA values of chicken samples, leading to r_P of 0.945 and RMSEC of 0.053 mg/100 g. The overall results showed that PLS model constructed with optimal wavelengths could be used to realize the rapid and contactless detection of TBA value in chicken.