RÉSUMÉ
<b>Objective::To investigate the feasibility of near-infrared spectroscopy for detecting the coating film thickness of Tianshu tablets. <b>Method::Nine batches of Tianshu tablets were taken during the coating process. Then, their near-infrared diffuse reflection spectra were collected. The sample set was divided into calibration set and validation set by Kennard-Stone algorithm. The preprocessing method was selected. The synergy interval partial least squares (siPLS) and moving window partial least squares (mwPLS) were employed to screen the optimal spectral interval. And the corresponding quantitative calibration model of partial least squares (PLS) were established. Some evaluation parameters were adopted to assess the performance of the model. <b>Result::The method of first derivative and Norris Derivative smoothing combined with standard normal variate transformation was suitable for processing the spectra. The optimal PLS model was established in the preferred band interval of siPLS. The correlation coefficient between the predicted value and the measured value of calibration set was 0.966, and the correlation coefficient between the predicted value and the measured value of validation set was 0.991.Both root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) values were small, which showed the fitness and predictive performance of the model were favorable. <b>Conclusion::The near-infrared spectroscopy technique can be used to determine coating film thickness of Tianshu tablets with high accuracy, which provides technical supports for the in-line determination of coating thickness in the production process of traditional Chinese medicine tablets.