摘要
多模态融合技术通过结合多源数据,可以克服单一模态的局限性。近年来,传感器以及遥感技术的发展为作物监测提供了更加丰富的数据源,光谱数据、图像数据、雷达数据以及热红外数据被广泛应用于作物监测中。通过利用计算机视觉技术以及数据分析方法,可以从中获取作物的表型参数、理化特征等信息,从而有助于评估作物的生长状况、指导农业生产管理。现有研究多数是基于单一模态数据展开,而单一模态的数据仅有一种类型的输入,缺乏对整体信息的理解,且容易受到单模态噪声的影响;部分研究虽然采用了多模态融合技术,但仍未能充分考虑模态间的复杂交互关系。为了深入分析多模态融合技术在农业领域应用的潜力,本文首先阐述了农业领域中多模态融合的先进技术与方法,重点梳理了多模态融合技术在作物识别、性状分析、产量预测、胁迫分析及病虫害诊断领域中的应用研究成果,分析了多模态融合技术在农业领域中存在的数据利用程度低、有效特征提取难、融合方式单一等问题,并对未来发展提出展望,以期通过多模态融合的方法推动农业精准管理、提高生产效率。
Multi-modal fusion technology,by combining data from multiple sources,has been widely applied in fields such as medicine,autonomous driving,and emotion recognition to overcome the limitations of a single modality.In recent years,advancements in sensor and remote sensing technologies have provided richer data sources for crop monitoring,including spectral data,image data,radar data,and thermal infrared data.By utilizing computer vision and data analysis methods,information such as phenotypic parameters and physicochemical characteristics of crops can be obtained,helping to assess crop growth and guide agricultural production management.Most existing studies were based on single-modal data,which involved only one type of input and lacked an understanding of the overall information,making them susceptible to noise from a single modality.Although some studies employed multi-modal fusion technology,they still did not fully consider the complex interactions between modalities.To thoroughly analyze the potential of multi-modal fusion technology in crop monitoring,the advanced technologies and methods of multi-modal fusion in the agricultural field were firstly outlined,with a focus on its application in crop identification,trait analysis,yield prediction,stress analysis,and pest and disease diagnosis.The existing challenges were also discussed and an outlook on future developments was provided,aiming to promote precision agriculture management and improve production efficiency through multi-modal fusion methods.
作者
李道亮
赵晔
杜壮壮
LI Daoliang;ZHAO Ye;DU Zhuangzhuang(College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China;National Innovation Center for Digital Fisheries,Beijing 100083,China)
出处
《农业机械学报》
北大核心
2025年第1期1-15,共15页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(32373186)。
关键词
多模态融合
传感器
遥感技术
作物监测
计算机视觉
农业精准管理
multi-modal fusion
sensors
remote sensing technology
crop monitoring
computer vision
precision agriculture management
作者简介
李道亮(1971-),男,教授,博士生导师,主要从事农业智能信息处理与农业农村信息化研究,E-mail:dliang@cau.edu.cn。
