FOOD SCIENCE ›› 2026, Vol. 47 ›› Issue (20): 31-36.doi: 10.7506/spkx1002-6630-20260420-199

• Food Analysis and Detection Based on Spectroscopy and Chemometrics • Previous Articles    

Rapid Nondestructive Detection of Rice Quality Based on Multi-source Data Fusion

LI Xue,, GUAN Qiwei, ZHANG Liangxiao, LIANG Shuilian, ZHAO Jie, XIAO Lu, LIU Wenwen, YANG Tusheng, WANG Xu,, CHEN Yan   

  1. 1. Institute of Quality Standard and Monitoring Technology for Agro-products of Guangdong Academy of Agricultural Sciences, Guangzhou 510640, China; 2. Guangdong Provincial Key Laboratory of Quality & Safety Risk Assessment for Agro-products, Guangzhou 510640, China; 3. Key Laboratory of Testing and Evaluation for Agro-product Safety and Quality, Ministry of Agriculture and Rural Affairs, Guangzhou 510640, China; 4. Oil Crops Research Institute, Chinese Academy of Agricultural Sciences, Wuhan 430062, China
  • Published:2026-11-06

Abstract: In this study, near infrared spectroscopy (NIR) and hyperspectral imaging (HSI) were used to determine the protein, amylose and moisture contents in rice. Two rice quality prediction models were developed using partial least squares (PLS) and support vector regression (SVR) based on the full-wavelength spectra and variable selection using competitive adaptive reweighted sampling (CARS). The prediction performances of the model based on the NIR, HSI and their fused data were comparatively analyzed. The results showed that the PLS model established by decision-level data fusion based on the full-wavelength spectra exhibited the highest performance in predicting the protein content of rice, with a coefficient of determination of prediction (R2 p) of 0.894 0, a root mean square error of prediction (RMSEP) of 0.287 3, and a ratio of performance to deviation (RPD) of 3.109 4. The SVR model constructed by CARS variable selection 收稿日期:2026-04-20 基金项目:广东省科学技术协会青年科技人才培育计划项目(SKXRC2025514); 广东省重点领域研发计划项目(2025B0202130004);广东省农业科学院现代种业创新能力提升工程项目(2026ZYTS14) 第一作者简介:李雪(1993—)(ORCID: 0009-0002-9713-1770),女,助理研究员,博士,研究方向为食品无损检测。 E-mail: 15071198582@163.com *通信作者简介:王旭(1981—)(ORCID: 0000-0002-4617-0860),女,研究员,博士,研究方向为农产品质量安全。 E-mail: wangxuguangzhou@126.com 陈岩(1982—)(ORCID: 0000-0003-0187-5020),男,研究员,博士,研究方向为农产品质量安全。 E-mail: chenyan@gdaas.cn combined with feature-level data fusion was the most effective for predicting the amylose content of rice (R2 p = 0.803 7, RMSEP = 2.267 9, RPD = 2.284 4). The SVR model developed based on the full-wavelength HSI data exhibited the best performance for predicting the moisture content of rice (R2 p = 0.894 8, RMSEP = 0.273 5, RPD = 3.121 1). Therefore, this method can be used for the rapid and non-destructive determination of protein, amylose, and moisture in rice, providing reliable technical support for the online classification and quality evaluation of rice.

Key words: rice; quality; near infrared spectroscopy; hyperspectral imaging; data fusion strategies

CLC Number: