您好，欢迎访问中国水产科学研究院 机构知识库！
中国水产科学研究院机构知识库

全部

  • 全部
  • 标题
  • 学者
  • 机构
  • 关键词

Fusion of near-infrared and Raman spectroscopy with machine learning strategies: Non-destructive rapid assessment of freshness and TVB-N value prediction in Pacific white shrimp (Litopenaeus vannamei)

文献类型： 外文期刊

作者： Tian, Zhenxing 1 ; Wu, Yanyan 1 ; Wei, Ya 1 ; Zhao, Yongqiang 1 ; Pan, Chuang 1 ; Wang, Yueqi 1 ;

作者机构： 1.Chinese Acad Fishery Sci, South China Sea Fisheries Res Inst, Natl R&D Ctr Aquat Prod Proc, Key Lab Aquat Prod Proc,Minist Agr & Rural Affairs, Guangzhou 510300, Peoples R China

2.Beibu Gulf Univ, Guangxi Coll, Coll Food Engn, Qinzhou 535000, Guangxi, Peoples R China

3.Beibu Gulf Univ, Univ Key Lab Dev & High Value Utilizat Buibu Gulf, Coll Food Engn, Qinzhou 535000, Guangxi, Peoples R China

4.Ocean Univ China, Coll Food Sci & Engn, Qingdao 266000, Peoples R China

关键词： Litopenaeus vannamei; Total volatile basic nitrogen; Spectral analysis machine learning; Data fusion; Non-destructive detection

期刊名称：FOOD RESEARCH INTERNATIONAL （ 影响因子：8.0； 五年影响因子：8.5 ）

ISSN： 0963-9969

年卷期： 2025 年 214 卷

页码：

收录情况： SCI

摘要： Total volatile base nitrogen (TVB-N) is a key indicator of shrimp freshness. Nevertheless, traditional detection methods are cumbersome, time-intensive, and destructive. Here, a rapid and non-destructive method based on near-infrared (NIR) and Raman spectroscopy for the assessment of TVB-N content in Litopenaeus vannamei was proposed. A TVB-N content prediction model was constructed based on three machine learning methods (Convolutional Neural Network, Extreme Learning Machine, and Backpropagation) combined with low-level and midlevel data fusion strategies. After Savitzky-Golay (SG) smoothing preprocessing, the NIR model with SPA feature extraction (coefficient of determination for prediction, R2p = 0.864) outperformed the Raman model with GA feature extraction (R2p = 0.784), with both being the optimal feature-level prediction models for their respective spectra. Furthermore, the combination of mid-level data fusion strategy and the Extreme Learning Machine model resulted in the best prediction performance, with R2p and root mean square error of prediction (RMSEP) values of 0.986 and 0.677 mg/100 g, respectively. Additionally, the feature-level fusion models optimized by the feature selection algorithm showed R2 values all exceeding 0.96. These results demonstrate the complementary advantages of NIR and Raman spectroscopy for non-destructive, real-time freshness monitoring of shrimp using portable instruments.

  • 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：中国水产科学研究院

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