Rapid detection of safflower quality based on hyperspectral technology

There are problems with quality inspection of safflower, high costs, and time-consuming quality inspection. Therefore, this study established a rapid model for the quality assessment of safflowers based on hyperspectral technology. In this study, the S-G CS, MSC, and FD methods were used to preprocess the hyperspectral data of the safflower. The feature wavelengths were extracted using SPA. Linear least squares regression (LSR) and nonlinear Levenberg-Marquardt algorithm (LMA) models were established for the hyperspectral prediction of HSYA, kaempferol, red pigment, and moisture content in safflower. The results showed that FD was used as the preprocessing method of spectra. The feature wavelengths were optimized by retaining the key wavelengths and merging similar wavelengths. A total of 15 wavelengths were identified as the feature wavelengths. The better model for predicting HSYA, kaempferol, red pigment, and moisture content was LMA, with r values of 0.96717, 0.94403, 0.97618, and 0.98090. In conclusion, the rapid detection model of safflower can accurately predict the contents of HSYA, kaempferol, red pigment, and moisture content. It can be used to control safflower quality during harvest, quality grading, and production processes. • Hyperspectral were used to construct physical and chemical models of safflower • S-G smoothing, MSC, FD and SD using data fusion with different preprocessing methods were developed • Safflowers with uneven qualities and various origins were modeled.