Abstract: The contents of lactic acid bacteria is an important indicator to evaluate the quality of chilled chicken breast. With the increasing of storage days the contents of lactic acid bacteria exceeds 10⁶ CFU/g, the chilled chicken breast becomes sticky and begins to rot. In order to predict the contents of lactic acid bacteria in chicken breast in his paper, the hyperspectral data were analyzed through chemometric algorithms. First of all, 119 samples of hyperspectral images of cold fresh chicken breast in range of 900~1700 nm were obtained, and the spectral information within the region of interests (ROIs) of the images were extracted. The original spectral data were pretreated by eight pretreatment methods, and partial least squares (PLS) algorithm was used for mining the spectral information to build F-PLS model in the full wavelength range. Then, regression coefficient method (RC), stepwise and successive projections algorithm (SPA) were applied for screening optimal wavelengths to optimize the F-PLS model. The results showed that the SPA-PLS model based on 21 optimal wavelengths (903.8, 905.5, 912.1, 915.4, 917.0, 920.3, 923.6, 931.8, 941.7, 1107.0, 1135.9, 1157.3, 1269.2, 1303.7, 1320.2, 1348.2, 1551.1, 1676.9, 1686.9, 1695.1 and 1698.4 nm) selected by SPA from baseline correction (BC) pretreatment spectra had best performance (r_P=0.949, RMSEP=0.439lg CFU/g, RPD=2.787). The results show that it would be feasible to predict the content of lactic acid bacteria in chilled chicken breast based on near-infrared hyperspectral imaging technology.