人工智能在海产品检测中的应用研究

Research on the application of artificial intelligence in seafood detection

中文关键词:  海产品检测  人工智能  光谱技术  无损检测  深度学习  数据融合

英文关键词:seafood detection  artificial intelligence  spectroscopic technology  non-destructive testing  deep learning  data fusion

中文摘要:

      随着全球海产品贸易的蓬勃发展与供应链的复杂化, 海产品品质与安全的快速、无损检测需求不断增加, 但真实海产品基质中的水峰吸收、米氏散射、内源荧光、峰重叠和批次差异会降低传统线性模型的稳定性。本文围绕拉曼光谱及表面增强拉曼散射(surface-enhanced Raman scattering, SERS)、中红外光谱(mid-infrared spectroscopy, MIR)、近红外光谱(near-infrared spectroscopy, NIR)、太赫兹光谱(Terahertz spectroscopy, THz)、紫外-可见吸收光谱(ultraviolet-visible spectroscopy, UV-Vis)和分子荧光光谱(molecular fluorescence spectroscopy, MFS)以及高光谱成像(hyperspectral imaging, HSI)等光谱与成像技术, 综述人工智能(artificial intelligence, AI)在海产品鲜度评价、物种真实性鉴别、冻融损伤识别、残留/毒素检测和微生物风险筛查中的应用, 重点归纳其在非线性建模、复杂模式识别和空间信息提取中的作用, 并讨论模型泛化、轻量化部署和标准方法确证等问题。本文可为海产品品质与安全快速筛查中光谱技术和AI模型的合理选择、性能评价及实际应用提供参考。

英文摘要:

      As global seafood trade continues to expand and supply chains become increasingly complex, there is a growing need for rapid and non-destructive methods to assess seafood quality and safety. However, water absorption, Mie scattering, endogenous fluorescence, overlapping spectral features and batch-to-batch variation in real seafood matrices can compromise the robustness of conventional linear models. This review examined a range of spectroscopic and imaging techniques, including Raman spectroscopy and surface-enhanced Raman scattering (SERS), mid-infrared spectroscopy (MIR), near-infrared spectroscopy (NIR), Terahertz spectroscopy (THz), ultraviolet-visible absorption spectroscopy (UV-Vis), molecular fluorescence spectroscopy (MFS), and hyperspectral imaging (HSI). It summarized the applications of artificial intelligence (AI) in freshness assessment, species authentication, detection of freeze-thaw damage, screening for residues and toxins, and microbial risk assessment in seafood. It focused on the roles of AI in nonlinear modeling, complex pattern recognition and spatial information extraction, and discussed issues such as model generalizability, lightweight deployment and confirmation by standard methods. This paper can serve as a reference for the rational selection, performance evaluation and practical application of spectroscopic techniques and AI models in the rapid screening of seafood quality and safety.