Abstract: The objectives of this research were to develop a hyperspectral imaging system to differentiate between fresh and frozen- thawed beef fillets.A total of 45 beef fillets were randomly divided into two groups,including 25 fresh and 20 frozen- thawed samples.In this study,spectral imaging dates were obtained by hyperspectral imaging system.All images were calibrated for reflectance,followed by segmentation to obtain the region of interest( ROI),and then the average spectral data was generated from the ROI images.Through multiplicative scatter correction( MSC) pretreatment to hyperspectral imaging data,optimal model of this test was built in range of full spectrum( 950~1500 nm) by means of partial least square regression( PLSR).The results showed that,the model had high prediction accuracy,the correct classification rate was 94.4%,and the near- infrared hyperspectral imaging technology was suit for recognizing to frozen- thawed beef.