Foundation model for generalist remote sensing intelligence:Potentials and prospects  

基础模型在通用遥感智能中的潜力与前景

作　　者：Mi Zhang Bingnan Yang Xiangyun Hu Jianya Gong Zuxun Zhang 张觅;杨炳楠;胡翔云;龚健雅;张祖勋

机构地区：[1]School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China [2]State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan 430079,China [3]Hubei Luojia Laboratory,Wuhan 430079,China

出　　处：《Science Bulletin》2024年第23期3652-3656,共5页科学通报(英文版)

基　　金：supported by the Key Research and Development Program of Hubei Province(2023BAB173);the State Key Laboratory of Geo-Information Engineering(SKLGIE2021-M-3-1);the National Natural Science Foundation of China(41901265);Major Program of the National Natural Science Foundation of China(92038301);supported in part by the Special Fund of Hubei Luojia Laboratory(220100028).

摘　　要：With the advent of Earth observation satellites,the remote sensing(RS)dataset has experienced exponential growth,significantly enhancing scientific research and applications.By early 2024,the global Earth observation constellation comprises 1,379 satellites,with projections indicating an increase to 5,500 by 2033.On a daily basis,these satellites produce more than 20 TB of raw data,leading to an accumulation exceeding 500 PB[1].The surge in data volume poses challenges in storage,analysis,and management within the remote sensing domain.Foundation models like ChatGPT,SAM,and CLIP[2],present novel approaches that improve efficiency and drive innovation in remote sensing data processing.Leveraging extensive training datasets,these models demonstrate promise across a range of remote sensing tasks[3–5].

关 键 词：FOUNDATION processing. EARTH 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]