Combining dual-wavelength laser-induced fluorescence hyperspectral imaging with mutual information decomposition and redundancy elimination method to detect Aflatoxin B1 of individual maize kernels

文献类型： 外文期刊

第一作者： Fan, Yaoyao

作者： Fan, Yaoyao;Kang, Jian;Chen, Liping;Fan, Yaoyao;Yao, Xueying;Wang, Zheli;Long, Yuan;Chen, Liping;Huang, Wenqian;Tian, Xi;Tian, Xi

作者机构：

关键词： Dual-wavelength; Fluorescence hyperspectral imaging; Mutual information; Information decomposition; Maize kernels; Aflatoxin B1

期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE （ 影响因子：8.9； 五年影响因子：9.3 ）

ISSN： 0168-1699

年卷期： 2025 年 239 卷

页码：

收录情况： SCI

摘要： Aflatoxin B1 (AFB1) contamination in maize kernels poses a serious threat to food safety. As a fluorescent compound, AFB1 can theoretically be detected under 365 nm excitation; however, endogenous fluorescent substances in maize interfere with its signal, resulting in low detection accuracy when using a single excitation wavelength. To address this challenge, we developed a dual-wavelength laser-induced fluorescence hyper-spectral imaging (DW-LIF-HSI) system that combines the high sensitivity of 365 nm excitation with complementary information from 340 nm excitation to mitigate spectral crosstalk. To fully exploit the dual-wavelength spectral data, a Mutual Information Decomposition and Redundancy Elimination (MIDRE) strategy based on the Synergistic-Unique-Redundant Decomposition (SURD) framework was proposed to quantify the unique, redundant, and synergistic information in the spectra and to select optimal feature wavelengths and wavelength pairs. These selected features were then used in a Classification-Prior-Guided Regression (CPGR) model, which incorporates predicted classification results as constraints to improve the accuracy of AFB1 quantification. Experimental results demonstrated that the MIDRE-based feature selection, combined with the CPGR model, significantly enhanced predictive performance, achieving an R2 of 0.8023, an RMSEP of 17.6560, and an RPD of 2.2491. Overall, the integration of DW-LIF-HSI with MIDRE provides a powerful and nondestructive approach for accurate AFB1 detection in maize kernels, with strong potential for applications in food safety monitoring and quality control.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序