A calibration model for determining the cotton content in cotton-terylene textile precisely was established.The near infrared spectra of 50 cotton-terylene textile samples were collected using self-designed accessory,and Savitsky-Goaly filter smoothing,first order derivative and second order derivative were used to pretreat the spectra.Calibration models for cotton content were established with Partial Least Square(PLS).The experimental results indicate that Savitsky-Goaly filter smoothing spectral treatments can not affect the performance of the model in this case,but NIR technique can analyze the cotton content in the textile precisely.The Root Mean Square Error of Calibration(RMSEC) and Root Mean Square Error of Prediction(RMSEP) are 0.022 and 0.018 respectively,and the error of analysis is restricted in ±0.05.In addition,several key aspects of this application were expressed.