多源数据驱动的蔬菜品质无损检测研究进展OA
Advances in Multi-source Data-Driven Nondestructive Assessment of Vegetable Quality
蔬菜品质无损检测是智慧农业与食品安全领域的重点研究方向.传统检测方法效率低、破坏性强,难以满足现代农业生产对快速、精准品质控制的需求.近年来,多源数据驱动的无损检测技术通过融合图像、光谱、声学、振动及仿生感官等多类信息,实现了对蔬菜外观、内部成分、缺陷及风味等多维品质的全面表征,显著提升了检测的准确性与智能化水平.本文系统综述了可见光成像、高光谱成像、近红外光谱、拉曼光谱与荧光光谱、声学振动传感、仿生感官数据等数据形态在各类蔬菜品质检测中的研究进展,分析了单源数据的优势与局限,并探讨了多源数据融合在提升模型鲁棒性方面的潜力与挑战.尽管多源数据技术已在实验室层面取得显著成果,但仍存在模型泛化能力不足、设备成本高、数据融合算法复杂等问题.未来研究应聚焦于跨模态特征融合、大规模数据库构建、轻量化传感器集成及全链条品质追溯系统开发,以推动多源数据无损检测技术在实际农业生产中的广泛应用.
Nondestructive assessment of vegetable quality is a priority research area in smart agriculture and food safety.Conventional analytical methods are inefficient and destructive,and thus fall short of modern demands for rapid and precise quality control.In recent years,multisource,data-driven nondestructive technologies,integrating imaging,spectroscopy,acoustics,vibration,and biomimetic sensory modalities,have enabled comprehensive characterization of multidimensional quality attributes,including appearance,internal composition,defects,and flavor,markedly improving accuracy and intelligence.This review systematically surveys advances in visible imaging,hyperspectral imaging,near-infrared spectroscopy,Raman spectroscopy and fluorescence spectroscopy,acoustic-vibration sensing,and bionic sensory data for vegetable quality assessment.It analyzes the strengths and limitations of single-source data and discusses the potential and challenges of multisource data fusion for enhancing model robustness.Despite notable laboratory-level progress,persistent issues remain,including limited model generalization,high instrumentation costs,and the algorithmic complexity of data fusion.Future work should prioritize cross-modal feature fusion,construction of large-scale databases,integration of lightweight sensors,and development of end-to-end quality traceability systems to enable widespread adoption of multisource nondestructive technologies in real-world agricultural production.
庄绿涵;郝浩源;奚嘉艺;王作辉;樊瑶瑶;黄文倩
北京市农林科学院智能装备技术研究中心,北京 100097||海南大学机电工程学院,海南海口 570228北京市农林科学院智能装备技术研究中心,北京 100097北京市农林科学院智能装备技术研究中心,北京 100097北京市农林科学院智能装备技术研究中心,北京 100097北京市农林科学院智能装备技术研究中心,北京 100097北京市农林科学院智能装备技术研究中心,北京 100097
农业科技
蔬菜品质无损检测图像数据光谱数据仿生感官数据多源数据融合
vegetable qualitynondestructive assessmentimaging dataspectral databiomimetic sensory datamultisource data fusion
《蔬菜》 2026 (8)
58-80,23
国家重点研发计划项目(2022YFD2002202)青年北京学者计划项目(2024-078).
评论