您好，欢迎访问中国农业科学院 机构知识库！

全部

  • 全部
  • 标题
  • 学者
  • 机构
  • 关键词

Fluorescence spectroscopy combined with a multi-task deep learning model for rapeseed oil quality analysis

文献类型： 外文期刊

作者： Jiang, Xinna 1 ; Liu, Quancheng 1 ; Cao, Xingda 1 ; Wang, Fan 1 ; Li, Long 4 ; Yan, Lei 1 ; Fan, Shuxiang 1 ;

作者机构： 1.Key Lab Natl Forestry ad Grassland Adm Forestry Eq, Beijing 100083, Peoples R China

2.State Key Lab Efficient Prod Forest Resources, Beijing, Peoples R China

3.Beijing Forestry Univ, Sch Technol, Beijing 100083, Peoples R China

4.Chinese Acad Agr Sci, Inst Food Sci & Technol, Key Lab Agroprod Proc, Minist Agr & Rural Affairs, Beijing 100083, Peoples R China

关键词： CNN; Multi-task deep learning; Quantitative analysis; Rapeseed oil; SHAP values

期刊名称：JOURNAL OF FOOD COMPOSITION AND ANALYSIS （ 影响因子：4.6； 五年影响因子：4.6 ）

ISSN： 0889-1575

年卷期： 2025 年 146 卷

页码：

收录情况： SCI

摘要： This study aims to develop a non-destructive, multi-task deep learning model for the simultaneous prediction of multiple rapeseed oil quality indicators based on ultraviolet fluorescence spectroscopy. A novel multi-task convolutional neural network (MT-CNN) architecture was constructed, integrating multi-head attention and residual connections to enhance feature extraction and gradient propagation. Five common spectral preprocessing techniques were evaluated, with the combination of standard normal variate (SNV) and Savitzky-Golay (SG) filtering yielding optimal results. The proposed model was further interpreted using Shapley Additive Explanations (SHAP) to identify key wavelengths contributing to the prediction of acid value, peroxide value, alpha-tocopherol, and polyphenols. To comprehensively evaluate modeling performance, this study compared partial least squares regression (PLSR), K-nearest neighbors (KNN), support vector regression (SVR), long shortterm memory (LSTM), and convolutional neural networks (CNN) under both single-task and multi-task learning frameworks. The results demonstrated that the MTDL model consistently outperformed all other models across all indicators, achieving an average correlation coefficient (R) of 0.9639. This study represents the first successful application of fluorescence spectroscopy with a multi-task deep learning model for rapeseed oil quality detection. This approach offers a rapid, accurate approach for simultaneous multi-parameter analysis of rapeseed oil constituents.

  • 相关文献
作者其他论文 更多>>
置顶
购物车
反馈

© 京ICP备09089781号-29 主办单位：中国农业科学院

学术资源： 中国农业科学院 农业专业知识服务系统 农科联盟资源共建共享平台 中国农业科技文献与信息服务平台 中国农业期刊集成服务平台