海洋水色遥感机器学习模型应用现状及定标与真实性检验AI模型展望  

CURRENT APPLICATIONS OF OCEAN COLOR REMOTE SENSING MACHINE LEARNING MODELS AND PROSPECTS FOR CALIBRATION AND VALIDATION

作　　者：陈鹏 厉运周 赵志刚 张思琪 张镇华 王军成 潘德炉 CHEN Peng;LI Yun-Zhou;ZHAO Zhi-Gang;ZHANG Si-Qi;ZHANG Zhen-Hua;WANG Jun-Cheng;PAN De-Lu(Qilu University of Technology(Shandong Academy of Sciences),Qingdao 266061,China;Laoshan Laboratory,Qingdao 266237,China;Second Institute of Oceanography,Ministry of Natural Resources,Hangzhou 310012,China)

机构地区：[1]齐鲁工业大学(山东省科学院),山东青岛266061 [2]崂山实验室,山东青岛266237 [3]自然资源部第二海洋研究所,浙江杭州310012

出　　处：《海洋与湖沼》2025年第1期3-24,共22页Oceanologia Et Limnologia Sinica

基　　金：国家自然科学基金,42322606号,42276180号,61991453号;山东省重点研发计划,2023ZLYS01号;国家重点研发计划项目,2022YFB3901703号,2022YFB3902603号,2022YFC3104200号;南方海洋科学与工程广东实验室引进人才重点专项,GML2021GD0809号。

摘　　要：海洋水色遥感技术是观测海洋水色状况的高效工具,数十年来持续提供了全球海洋光学特性、水色和生物地球化学参数的重要数据。随着人工智能的快速发展,基于机器学习模型的水色遥感研究逐渐成为科研领域的新热点。本文回顾了海洋水色遥感中机器学习模型的现状与挑战,并评估了这些模型在大气校正、水色反演、碳循环及数据重构中的应用效果。本文重点综述了机器学习在海洋水色遥感应用中的进展。鉴于卫星传感器长期在轨运行中,可能遭受元器件老化等问题,本文强调了对遥感器进行持续定标和真实性检验的必要性。这些检验工作是确保卫星遥感数据质量的基础,对于卫星遥感应用至关重要。因此,本文还展望了未来在水色遥感定标与真实性检验中,人工智能大模型的发展前景。As an effective means for rapid,large-scale,and continuous observation of ocean color conditions,ocean color remote sensing technology has provided a global view of ocean optical properties,water color,and biogeochemical parameter distributions for several decades.In recent years,with the advancement of artificial intelligence technologies,research on ocean color remote sensing based on machine learning models has become a burgeoning field.This paper aims to review the current applications and challenges of machine learning models in ocean color remote sensing.It analyzes the performance of various algorithms in atmospheric correction,water color inversion,carbon cycle studies,and data reconstruction.Additionally,it discusses the considerations for using machine learning models in ocean color remote sensing,focusing particularly on their application progress in this domain.Given that satellite sensors are subject to degradation over long-term operations,continuous calibration and validation are essential to ensure the quality of observational data.Calibration and validation are fundamental to the application of satellite remote sensing.Therefore,this article also explores the future prospects of large-scale artificial intelligence models in the calibration and validation of ocean color remote sensing.

关 键 词：水色遥感 大模型 机器学习 辐射定标 真实性检验 大气校正 

分 类 号：P731[天文地球—海洋科学]