Quantitative detection of fatty acid value in brown rice using hyperspectral imaging combined with chemometric methods

• Published in: Infrared physics & technology Vol. 150; p. 106049
• Main Authors: Cai, Ting, Liu, Xingquan, Li, Qianqian, Yang, Dong, Jie, Yu, Shi, Tianyu
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.11.2025
• Subjects: Brown rice; Chemometrics; Fatty acid value; Hyperspectral imaging; Quantitative detection; Brown rice; Fatty acid value; Chemometrics; Quantitative detection; Hyperspectral imaging
• ISSN: 1350-4495
• DOI: 10.1016/j.infrared.2025.106049

•HSI and chemometrics were combined for detecting fatty acid value of brown rice.•The wavelengths of sample with different particle sizes were extracted by SPA.•PLSR and SVR were developed based on different spectra features.•The best BOC-SPA-SVR model yielded satisfactory performances. Fatty acid value was one of the important indexes to evaluate brown rice quality during storage. In this study, brown rice (Dao Huaxiang No. 2) was chosen as the experimental subject. Three different particle sizes of brown rice powder samples were explored to the change rule of the visible near infrared hyperspectral image and fatty acid value. Detection models for the fatty acid value of brown rice powder samples were established based on the full and feature spectra. Five pre-processing methods(including baseline offset correction, et al.) were used to process the full spectral data and the successive projections algorithm (SPA) was utilized to select feature wavelengths. And the partial least squares regression (PLSR) and the support vector regression (SVR) models were established based on both full wavelengths and feature wavelengths. The comparative results indicated that the BOC-SPA-SVR model had the highest performance, with a determination coefficient (Rp2) of 0.9954, root mean square errors (RMSEP) of 0.6693, and residual prediction deviation (RPD) of 10.4357 in the prediction set. The results demonstrated that HSI with BOC-SPA-SVR model performed well in the detection of fatty acid value of the brown rice, and the proposed BOC-SPA-SVR model in this study can be used for the development of the optical rapid detection equipment for brown rice’s fatty acid value and other quality indicators.