OBJECTIVE To establish a quantitative calibration model for total ash content in Croci Stigma by using near infrared diffuse reflection spectrometry(NIRS), and to provide a technical foundation for its rapid and non-destructive detection.

METHODS A total of 220 batches of Croci Stigma samples were used to determine total ash content and acquire NIR spectra. Multiple spectral preprocessing methods, including standard normal variate(SNV), multiplicative scatter correction(MSC), detrending(DT), Savitzky-Golay smoothing, Savitzky-Golay derivatives, and mean centering, were applied to the raw spectra. A quantitative calibration model for total ash was developed using partial least squares(PLS) regression. The performance of models under different preprocessing combinations was compared to identify the optimal strategy. The predictive accuracy of the model was further validated using the test set.

RESULTS Both the combination of SNV+DT+Savitzky-Golay smoothing+Savitzky-Golay derivatives+mean centering and the combination of MSC+Savitzky-Golay smoothing+Savitzky-Golay derivatives+mean centering were the optimal spectral preprocessing methods. Based on these approaches, the near infrared calibration models developed for total ash in Croci Stigma demonstrated good predictive performance.

CONCLUSION The developed near infrared calibration models allow for rapid, non-destructive, and accurate determination of total ash in Croci Stigma, making it suitable for real-time quality control and analysis of Croci Stigma.