摘要
随着人工智能与遥感领域技术的深度融合,遥感大模型逐渐成为当前的研究热点.本文系统梳理了遥感大模型的发展历程和最新进展,分别从模态和任务的角度分类总结了遥感大模型关键技术新动态.从模态角度,遥感大模型在处理海量遥感数据时展现出卓越的能力,可有效挖掘多模态遥感大数据中复杂的空间和光谱信息.从任务角度,遥感大模型正从单任务向多任务处理演变,同时展现出强大的泛化能力,具备迅速适应多模态数据环境下的多样化任务需求特性.首先,归纳总结了单任务/多任务、单模态/多模态遥感大模型学术界热点研究;其次,分类梳理了农业大模型实际应用现状;最后,结合遥感大模型在农业领域泛化能力和可用性等方面关键科学问题进行展望,并重点聚焦于农业知识图谱构建、数据迁移以及轻量化部署三个方面的分析.
As artificial intelligence and remote sensing technologies converge,foundation models in remote sensing have emerged as a major research focus.This paper systematically reviews the evolution and recent breakthroughs in remote sensing foundation models,classifying and analyzing key techniques along modality and task dimen-sions.From the modality perspective,remote sensing foundation models excel in processing large-scale remote sensing data,efficiently extracting intricate spatial-spectral features from multimodal data streams.Task-wise,remote sensing foundation models are transitioning from single-task to multi-task paradigms,demonstrating ro-bust generalization that enables rapid adaptation to varied tasks in multi-modal data contexts.Firstly,this paper systematically reviews cutting-edge research on single/multi-task and single/multi-modal remote sensing founda-tion models,then examines agricultural foundation model implementations,and finally forecasts critical scientific challenges in enhancing the generalization and applicability of remote sensing foundation models for agricultural applications.Furthermore,this study focuses on the analysis of agricultural knowledge graph construction,data migration,and lightweight deployment.
作者
钱育蓉
白璐
刘鹏
李晨
杨帆
李梦倩
范迎迎
刘炫辰
公维军
QIAN Yurong;BAI Lu;LIU Peng;LI Chen;YANG Fan;LI Mengqian;FAN Yingying;LIU Xuanchen;GONG Weijun(School of Computer Science and Technology,Xinjiang University,Urumqi Xinjiang 830017,China;Xinjiang Engineering Research Center of Big Data and Intelligent Software,School of Software,Xinjiang University,Urumqi Xinjiang 830091,China;Xinjiang Research Institute of the Huairou Laboratory,Urumqi Xinjiang 830013,China;School of Information Management,Xinjiang University of Finance and Economics,Urumqi Xinjiang 830012,China;School of Information Technology and Communication,Hexi University,Zhangye Gansu 734000,China)
出处
《新疆大学学报(自然科学版中英文)》
2025年第4期401-415,共15页
Journal of Xinjiang University(Natural Science Edition in Chinese and English)
基金
怀柔实验室项目(YZD2024025A)
新疆维吾尔自治区杰出青年科学基金“基于多时相协同变化检测的基本农田遥感监测研究”(2023D01E01)
国家自然科学基金“面向数字孪生农业的空-天-地多模态数据融合关键技术研究”(62266043)
新疆维吾尔自治区青年拔尖人才项目“基于多源遥感大数据的粮食安全关键技术研究”(2023TSYCCX0043)
新疆维吾尔自治区天山创新团队项目“面向农业的天地协同水资源时空精准调度研究及应用”(2023D14012)
新疆维吾尔自治区重点研发专项“边防异常信息多元感知与智能监测研究”(2023B01029-1,2023B01029-2)
新疆维吾尔自治区自然科学基金“基于深度学习的多源数据融合技术在新疆棉田识别与产量预测中的应用研究”(2022D01B123)
新疆维吾尔自治区研究生科研创新项目“面向复杂场景的遥感图像耕地变化检测研究”(XJ2025G036)
甘肃省科技计划自然科学基金“基于深度学习的复杂场景细粒度表情识别算法研究”(25JRRG037).
关键词
遥感大模型
单任务大模型
多任务大模型
农业遥感应用
remote sensing foundation model
single-task foundation model
multi-task foundation model
agri-cultural remote sensing applications
