Abstract: Abstract: This study aims to establish a rapid detective method for the characteristics of beef texture through near infrared spectroscopy. The diffuse reflectance spectra at 800-2500 nm of 202 fresh beef samples were collected. The hardness, springiness, chewiness and the adhesiveness of beef samples were tested by texture analyzer. After wavelet denoising, the data were collected through processing the diffuse reflectance spectra by smoothness, first order differential, second order differential, standard normal variate (SNV), SNV combined with first order differential or SNV combined with second order differential respectively, and then the partial least square regression (PLSR) statistical analysis models for the near-infrared spectra of beef samples were established. According to the predicted results of hardness, springiness and chewiness of beef through 6 PLSR statistical analysis models, the optimal predictive model was the PLSR statistical analysis model for the near-infrared spectra by the second order differential preconditioning method after wavelet denoising. The correlation coefficients of their calibration sets for hardness, springiness and chewiness of beef samples were all above 0.9, and the root mean square errors of calibration (RMSEC) were 6.247 N, 0.760 mm and 14.954 mJ, respectively. The correlation coefficients of their prediction sets were all above 0.664, and the root mean square errors of prediction (RMSEP) were 8.887 N, 0.951 mm and 22.117 mJ, respectively. The ratio of prediction to deviation (RPD) were 2.43, 1.88 and 2.32, respectively. The high predictive accuracy of hardness and chewiness of beef samples was obtained through the established optimal model, which was also fit for predicting the springiness of beef with low accuracy. The hardness, springiness and chewiness of beef are closely related to its own water, protein, fat and other chemical components. These chemical components of beef can be detected using near infrared spectra. Therefore, the hardness, springiness and chewiness of beef can be predicted through near infrared spectroscopy technology. The correlation coefficients of the adhesiveness in the calibration set and prediction set were high (0.720 and 0.694), and RMSEC and RESEP were small (0.302 and 0.243 N·mm) by using first order differential preconditioning method after wavelet denoising, but its RPD value was less than 1.5, so the prediction performance of the detection model of beef adhesiveness was not very satisfactory and the model was not fit for predicting the adhesiveness of unknown samples. This may be due to the adhesiveness value of each sample was the average value of 4 points in different parts of the sample. The difference was relatively large between the measured values, and the performance of the prediction model was affected by the accuracy of the measurement results. Further studies were needed for establishing the accuracy predictive model of the adhesiveness of beef by near infrared spectroscopy. The present study has established a rapid and accurate predictive model, which possesses the features of nondestructive inspection, qualitative analysis and quantitative detection for beef texture characteristics.