Abstract: Near-infrared hyperspectral imaging system ranging from 900 nm to 1700 nm was used to study chilled mutton tenderness detection intactly and rapidly. Shear force being the criterion of tenderness,collected the hyperspectral scattering image of the chilled mutton samples and extracted the reflectance spectrum at range of interest. Used original spectrum, characteristic region spectrum and Savitzky- Golay convolution smoothing preprocessing spectrum to establish partial least squares regression( PLSR) model of chilled mutton's tenderness,the model of pretreatment characteristics bands had better results. The result showed that characteristic region spectrum could effectively replace the whole band spectrum. After Savitzky-Golay convolution smoothing,model was better to predict,the correlation coefficient(RP) and the predict root mean square error(RMSEP) were 0.773 and 1.060. The research demonstrated that it was feasible to detect chilled mutton tenderness intactly and rapidly by near-infrared hyperspectral imaging technology associating with partial least squares regression method.