统一知识库检索

全局检索与知识浏览

跨论文、博客、数据集线索、项目和工具统一检索。检索结果可以继续跳转到独立问答页,做语义追问和来源核验。

检索条件
用当前检索词继续问答
当前展示 24 条,共 3235
筛选后的知识库结果
首页/全局检索
资讯
Felt Blog
Platform
UrbanComputing
什么是 GeoAI?人工智能如何增强地理信息系统(GIS)与地理空间数据分析
What’s GeoAI? How AI enhances GIS and geospatial data analytics

探讨 GeoAI 的基本原理,即人工智能(AI)与地理信息系统(GIS)相结合,用于分析空间数据、识别空间模式,并提供可操作的地理空间洞察。

Mamata Akella
2026/03/25
论文
arXiv
GeoAI
GIS
GeoAI 智能体基础单元
GeoAI Agency Primitives

本文介绍针对 GeoAI 助理的智能体基础单元(agency primitives)的持续研究——这些是将基础模型(Foundation models)与以人工制品(artifacts)为中心、以人为中心(human-in-the-loop)的工作流相连接的核心能力,而地理信息系统(GIS)从业者实际工作正发生于此类工作流中。尽管卫星影像字幕生成、视觉问答及可提示分割等技术已取得进展,但这些能力尚未为从业者带来实际生产力提升;后者大部分时间用于生成矢量图层、栅格地图和制图成果。这一差距不仅源于模型能力本身,更在于缺乏一个支持迭代协作的智能体层(agency layer)。我们为此类智能体层提出一套包含 9 个基础单元的术语体系,涵盖导航(navigation)、感知(perception)、地理参考记忆(geo-referenced memory)与双重建模(dual modeling)等,并配套设计了一项衡量人类生产力的基准测试。本研究的目标是构建一套术语体系,使 GIS 领域中的智能体辅助功能具备可实现性、可评测性与可比性。

Akram Zaytar, Rohan Sawahn, Caleb Robinson
2026/04/02
论文
Taylor & Francis
PaperDiscovery
全文:GeoFM:地理基础模型将如何重塑空间数据科学与GeoAI?
Full article: GeoFM: how will geo-foundation models reshape spatial data science and GeoAI?

本文简要概述了现有的地理基础模型(GeoFM)与地理人工智能(GeoAI)模型,以及用于评估这些模型的核心数据集与基准测试。地理基础模型仍是一个新兴且快速发展的研究领域。依据基础模型(FM)或地理基础模型(GeoFM)在各项研究中所起的作用,可将现有GeoFM相关研究大致分为以下三类:1)通过提示工程(prompt engineering)与任务特定微调(task-specific fine-tuning),将现有基础模型适配至地理空间任务;2)构建面向地理空间任务的先进大语言模型(LLM)智能体框架;3)通过具备地理感知能力的模型训练与微调,开发新型地理基础模型。

入库时间:2026/03/26
论文
基于手机数据的情境感知人口位移估计:一种方法论框架
PaperDiscovery
GeoAnalystBench:面向空间分析工作流与代码生成的地理人工智能(GeoAI)基准测试
GeoAnalystBench: A GeoAI benchmark for assessing large language models for spatial analysis workflow and code generation

近期,GeoBenchX [20] 和 GEOBench-VLM [7] 等基准测试拓展了该领域,致力于解决地理空间任务中的多步骤、多模态及基于工具的推理挑战。ScienceAgentBench [6] 覆盖地理信息科学等四个科学学科,共包含102项任务,并基于自包含的 Python 程序进行严格的执行导向评估,使其直接适用于地理信息系统(GIS)……

2025/09/07
项目
GitHub Repositories
GeoAI
geographic intelligent agent
MEKXH/golem: Golem — 面向地理空间行业的自演化 GeoAI 智能体 — 支持自然语言驱动的 GDAL/PostGIS 工作流、学习式流水线复用、生成式工具脚手架、审批与审计治理机制,并可通过 WebUI、TUI 及即时通讯(IM)渠道访问。
MEKXH/golem: Golem — The self-evolving GeoAI Agent for the geospatial industry — natural-language GDAL/PostGIS workflows, learned pipeline reuse, fabricated tool scaffolding, approval & audit governance, accessible via WebUI, TUI, and IM channels.

Golem — 面向地理空间行业的自演化 GeoAI 智能体 — 支持自然语言驱动的 GDAL/PostGIS 工作流、学习式流水线复用、生成式工具脚手架、审批与审计治理机制,并可通过 WebUI、TUI 及即时通讯(IM)渠道访问。本 GitHub 仓库由 MEKXH(开发者)维护。主要编程语言:Go。GitHub Star 数:165。最后更新时间:2026-03-29。

MEKXH
2026/03/29
论文
Wiley
PaperDiscovery
地球科学中的人工智能:GeoAI 视角
Artificial Intelligence in Earth Science: A GeoAI Perspective - Li - 2025 - Journal of Geophysical Research: Machine Learning and Computation - Wiley Online Library

这些约束条件为训练与验证 GeoAI 模型提供了独特机会,从而提升其地理可迁移性。上述要素共同凸显了 GeoAI 的独特性——它不仅是一种人工智能应用,更是一个专门应对地球科学中时空复杂性的学科领域。地球科学领域一项开创性的 GeoAI 建模工作是 Prithvi-EO 的开发;这是一种新型地理空间 AI 基础模型,基于具有国家及全球覆盖范围的时间序列遥感(EO)数据进行训练,旨在提取独特的语义与光谱特征。

2025/07/22
论文
ACM
PaperDiscovery
基础模型赋能地理人工智能(GeoAI)的机遇与挑战(展望论文)|ACM《空间算法与系统汇刊》
On the Opportunities and Challenges of Foundation Models for GeoAI (Vision Paper) | ACM Transactions on Spatial Algorithms and Systems

本文探讨了构建面向地理人工智能(GeoAI)的多模态基础模型(Foundation Models, FMs)所蕴含的潜力与面临的挑战。我们首先通过在七个任务上评估多个现有基础模型的性能,考察其在多个地理空间领域(包括地理空间语义、健康地理学、城市地理学和遥感)中的适用性。

入库时间:2026/03/26
论文
Wiley
PaperDiscovery
GeoAnalystBench:面向空间分析工作流与代码生成的 GeoAI 基准——评估大语言模型在空间数据分析中的能力
GeoAnalystBench: A GeoAI Benchmark for Assessing Large Language Models for Spatial Analysis Workflow and Code Generation - Zhang - 2025 - Transactions in GIS - Wiley Online Library

本研究通过覆盖多种应用场景,旨在评估当前最先进的大语言模型(LLM)在生成地理处理工作流及解决地理空间问题方案方面的能力,并揭示此类模型在空间数据科学自动化进程中日益增强的作用。为构建基准测试任务,我们综合整理了2025年之前公开可获取的GIS教程(例如Esri Learn平台、高校实验指导材料)中的地理空间问题求解实例。

2025/10/13
论文
Springer
PaperDiscovery
以人为中心的GeoAI基础模型:GeoAI与人类动态的交汇 | 城市信息学 | Springer Nature Link
Human-centered GeoAI foundation models: where GeoAI meets human dynamics | Urban Informatics | Springer Nature Link

本研究探讨人类动态在地理空间人工智能(GeoAI)中的作用,强调其重塑地理空间研究的潜力。

入库时间:2026/03/26
论文
International Journal of Geographical Information Science
PublisherJournal
GeoAI
地理人工智能研究的转型:主题结构、时间演化与方法-领域关联
GeoAI research in transition: thematic structures, temporal evolution and method–domain linkages
Usman Mehmood Uznir Ujang Suhaibah Azri a Faculty of Built Environment and Surveying, 3D City Modeling Lab, Universiti Teknologi Malaysia, Johor Bahru, Malaysiab No. 8 Geodetic Unit, Directorate of Printing and Geodesy, Survey of Pakistan, Rawalpindi, PakistanUsman Mehmood, PhD, is a geospatial researcher with a PhD in Geoinformatics from Universiti Teknologi Malaysia (UTM). His research interests include GIS, 3D city modelling, spatio-temporal analysis and the application of 3D GIS and machine learning for intelligent maintenance of high-rise buildings and smart urban systems. His contributions include conceptualization, data collection, analysis, methodology and writing.Uznir Ujang, PhD, is an Associate Professor at Universiti Teknologi Malaysia (UTM) in the Faculty of Built Environment and Surveying, where he leads the 3D GIS Research Group. His research interests include 3D city modelling, 3D GIS, topology and Geographic Information Science (GISc) for urban geospatial analysis. His contributions include conceptualization, methodology, supervision, writing, review and editing.Suhaibah Azri, PhD, is a Senior Lecturer in the Geoinformation Department, Faculty of Built Environment and Surveying, Universiti Teknologi Malaysia (UTM). Her research focuses on geospatial data organization and management, relational databases, and 3D GIS and spatial analysis for urban planning and smart city applications. Her contributions include conceptualization, writing, review and editing.
2026/01/26
论文
Annals of GIS
PublisherJournal
GeoAI
开源大语言模型在塑造地理人工智能未来中的作用
The role of open-source LLMs in shaping the future of GeoAI

大语言模型(LLMs)正在重塑地理空间人工智能(GeoAI),在数据处理、空间分析和决策支持方面带来新的能力。本文探讨了开源范式在此变革中的关键作用。尽管专有大语言模型具有易用性,但通常限制了定制化、互操作性和透明度,而这些对于专业地理空间任务至关重要。相反,开源替代方案通过促进更高的适应性、可复现性以及社区驱动的创新,显著推动了地理信息科学(GIScience)的发展。开放框架使研究人员能够定制解决方案,集成前沿方法(如强化学习、高级空间索引),并遵循FAIR(可发现、可访问、可互操作、可重用)原则。然而,对任何大语言模型日益增长的依赖,要求我们审慎考虑安全漏洞、伦理风险以及对人工智能生成地理空间成果的健全治理。本文认为,地理信息科学的进步不应依赖单一模型类型,而应通过培育一个多样化、可互操作的生态系统来实现,该系统结合开源基础以促进创新、定制化的地理空间模型以及跨学科协作。通过对开源大语言模型在更广泛GeoAI格局中机遇与挑战的批判性评估,本文为有效利用大语言模型推动空间研究、政策制定与决策的公平、可持续且科学严谨的发展,贡献了一场深入的学术讨论。

Xiao Huang Zhengzhong Tu Xinyue Ye Michael Goodchild a Department of Environmental Sciences, Emory University, Atlanta, GA, USAb Department of Computer Science and Engineering, Texas A&M University, College Station, TX, USAc Department of Landscape Architecture and Urban Planning & Center for Geospatial Sciences, Applications and Technology, Texas A&M University, College Station, TX, USAd Department of Geography, University of California, Santa Barbara, CA, USA
2026/02/13
论文
arXiv
GeoAI
GIS
评估GeoAI解释与遥感领域知识在卫星洪水制图中的对齐性
Evaluating the Alignment Between GeoAI Explanations and Domain Knowledge in Satellite-Based Flood Mapping

卫星数量的持续增加提升了地球观测的时间分辨率,使基于卫星的洪水制图成为业务化洪水监测中一种颇具前景的方法。作为地理空间人工智能(GeoAI)的重要应用,基于深度学习的卫星影像洪水制图方法通过从海量遥感数据中学习复杂的空谱模式,显著提升了预测性能。然而,深度学习模型决策过程的不透明性仍是其融入关键科学与业务工作流的主要障碍。这凸显了系统评估模型解释是否符合既定遥感领域知识的必要性。为填补这一研究空白,本研究提出了ADAGE(Alignment between Domain Knowledge And GeoAI Explanation Evaluation,领域知识与GeoAI解释对齐性评估)框架。该框架旨在系统评估深度学习模型解释与既定遥感知识(特别是地表独特光谱特性)之间的对齐程度。ADAGE框架采用通道分组SHAP(SHapley Additive exPlanations)方法,估算分组输入通道对像素级预测的贡献。在两项基于卫星的洪水制图任务上的实验表明,ADAGE框架能够:(1)定量评估模型解释与基于领域知识生成的参考解释之间的对齐性;(2)通过对齐得分辅助领域专家识别未对齐的解释。本研究有助于弥合GeoAI可解释性与地球观测领域知识之间的鸿沟,提升GeoAI模型的适用性。

Hyunho Lee, Wenwen Li
2026/04/29
论文
International Journal of Geographical Information Science
PublisherJournal
GeoAI
GeoAI的坍塌?合成地理空间数据使用带来的伦理问题
GeoAI collapse? Ethical implications of synthetic geospatial data use
Bo Zhao Yue Lin a Department of Geography, University of Washington, Seattle, WA, USAb Department of Geography and Geographic Information Science, University of Illinois Urbana-Champaign, Urbana, IL, USABo Zhao is a Professor in the Department of Geography at the University of Washington, Seattle. His research lies at the intersection of geospatial technologies and the humanities, with recent work focusing on the social implications of GeoAI and other emerging geospatial technologies, as well as fake geographies. He contributed to the conceptualization, main writing, and revisions of the paper.Yue Lin is an Assistant Professor in the Department of Geography & Geographic Information Science at the University of Illinois Urbana-Champaign. Her research interests include spatial data science, location privacy, and the societal implications of algorithmic systems. She contributed to the conceptualization and the design and execution of the experiments for this paper.
2026/01/06
论文
International Journal of Geographical Information Science
PublisherJournal
GeoAI
街道语义树:一种面向城市电动滑板车骑行量分类的知识驱动GeoAI框架
Street semantic tree: a knowledge-driven GeoAI framework for urban e-scooter ridership classification
Huihai Wang William Davis Yiming Xu Justin Yu Gengchen Mai Junfeng Jiao a Urban Information Lab, School of Architecture, The University of Texas at Austin, TX, USAb Urban Studies, The University of Texas at Austin, TX, USAc SEAI Lab, Department of Geography and the Environment, the University of Texas at Austin, TX, USAHuihai Wang is a PhD candidate in Community and Regional Planning program at the University of Texas at Austin, USA. His research focuses on geospatial AI, knowledge graphs, and urban intelligence for autonomous vehicles and delivery robots.William Davis is a Technical Program Management Analyst at Dell Technologies. He graduated from the University of Texas at Austin, where his academic work focuses on using data and geographic information systems to inform urban planning and decision-making.Yiming Xu is a Postdoctoral Fellow in Community and Regional Planning at the University of Texas at Austin, USA. He earned his Ph.D. in Civil Engineering from the University of Florida in 2023 and holds M.S. and B.E. degrees in Transportation Engineering from Tongji University in Shanghai, China. His research focuses on innovative mobility, travel behavior modeling, and transportation equity, combined with advanced artificial intelligence methodologies.Justin Yu is a student at Westwood High School in Austin, Texas, USA, and a research intern at the Urban Information Lab at the University of Texas at Austin, USA. His research interests focus on applying computer science and data-driven methods to study transportation systems, the environment, sustainability, equity, and smart cities.Gengchen Mai is an Assistant Professor at the Department of Geography and the Environment, the University of Texas at Austin, USA. He is also a visiting faculty researcher at Google Research. His research interests include geo-foundation models, geographic question answering, and spatially explicit artificial intelligence models.Junfeng Jiao is an associate professor and director of the Urban Information Lab at the University of Texas at Austin, USA. His research focuses on Smart City, Urban Informatics, and AI.
2025/10/29
论文
arXiv
GeoAI
GIS
受热力学启发的可解释地理人工智能:揭示异质空间系统中的状态依赖机制
Thermodynamic-Inspired Explainable GeoAI: Uncovering Regime-Dependent Mechanisms in Heterogeneous Spatial Systems

建模空间异质性及其相关临界转变,仍是地理学与环境科学中的基础性挑战。尽管传统的地理加权回归(GWR)和深度学习模型提升了预测能力,但它们往往难以阐明状态依赖的非线性关系——即驱动因子在不同异质区域中可能呈现相反的功能作用。我们提出一种受热力学启发的可解释地理空间人工智能框架,将统计力学与图神经网络相融合。该框架将空间变异性概念化为系统负荷(E)与容量(S)之间的热力学竞争,从而解耦驱动空间过程的潜在机制。我们在三个模拟数据集及三个跨领域真实数据集(住房市场、心理健康患病率、野火引发的PM2.5异常)上开展实验,结果表明,新框架成功识别出预测因子在不同状态下的角色反转现象,而标准基线模型均未能发现此类现象。值得注意的是,该框架明确诊断出2023年加拿大野火事件期间系统向负荷主导态的相变过程,从而将物理机制转变与统计异常区分开来。这些发现表明,引入热力学约束可在保持复杂空间系统强预测性能的同时,提升地理人工智能(GeoAI)的可解释性。

Sooyoung Lim, Zhenlong Li, Zi-Kui Liu
2026/04/06
论文
International Journal of Geographical Information Science
PublisherJournal
GeoAI
多尺度制图中地图综合的GeoAI:基础、研究议程与跨学科视角
GeoAI for map generalization in multi-scale cartography: foundations, a research agenda, and interdisciplinary perspectives
Zhiyong Zhou Cheng Fu Yu Feng Guillaume Touya Monika Sester Robert Weibel a Department of Geography, University of Zurich, Zurich, Switzerlandb Department of Geography, University of Wisconsin-Madison, Madison, USAc College of Information and Electrical Engineering, China Agricultural University, Beijing, Chinad i3mainz - Institute for Spatial Information and Surveying Technology, University of Applied Sciences Mainz, Mainz, Germanye Chair of Cartography and Visual Analytics, Technical University of Munich, Munich, Germanyf LASTIG, IGN, ENSG, University Gustave Eiffel, Champs-sur-Marne, Franceg Institute of Cartography and Geoinformatics, Leibniz University Hannover, Hannover, Germanyh UZH Healthy Longevity Center, University of Zurich, Zurich, SwitzerlandZhiyong Zhou is an SNSF postdoctoral research fellow at the Department of Geography, University of Wisconsin-Madison. Before this position, he was a postdoc at the Department of Geography, University of Zurich, from which he obtained a Ph.D. in Geography/Earth System Science in 2022. His research interests include GeoAI, location-based services (LBS), computational cartography, and urban analytics. He contributed to the writing of the original draft, conceptualization, methodology, investigation, and validation.Cheng Fu is a professor at China Agricultural University. His research interests include place modeling, human mobility, and map generalization. He contributed to review, editing, conceptualization, methodology, investigation, supervision, and funding acquisition.Yu Feng is a professor at Hochschule Mainz – University of Applied Sciences. Prior to his current position, he worked as a group leader and postdoctoral researcher at the Technical University of Munich (TUM). Earlier in his career, he was a Research Associate at Leibniz University Hannover, where he also completed his PhD. His research interests include GeoAI, volunteered geographic information (VGI), and cartographic generalization. He contributed to review, editing, conceptualization, methodology, and investigation.Guillaume Touya is a senior researcher, at IGN France (the French mapping agency) and Univ Gustave Eiffel, holding a PhD and habilitation in GI science from Paris-Est University. His research interests focus on automated cartography, map generalisation and volunteered geographic information, with a particular interest in research approaches to multi-scale cartography that mix automated cartography, spatial cognition and human–computer interaction issues. He is the principal investigator of the recent LostInZoom project, funded by the Europe Research Council (ERC) and the chair of the ICA (International Cartographic Association) commission on map generalisation and multiple representation. He contributed to review, editing, conceptualization, methodology, and investigation.Monika Sester is a surveying engineer by training (University Karlsruhe) and earned her PhD on a topic of Machine Learning at the University of Stuttgart in 1995, and her habilitation in 2000 on the automatic generation of multiple representations of geodata. Since 2000 she has been professor and head of the Institute of Cartography and Geoinformatics (ikg) at Leibniz University Hannover. She and her group work on the automation of spatial data processing with methods from computational geometry, optimization and AI. She contributed to review, editing, conceptualization, methodology, and investigation.Robert Weibel is a Professor Emeritus of Geographic Information Science at the University of Zurich, Switzerland. He is interested in computational cartography; mobility analytics with applications in health, transportation and animal ecology; and spatial analysis and modeling for the study of language evolution. From 1992 he built up the ICA Commission on Map Generalization and served as Commission Chair until 2003. He contributed to reviewing, editing, conceptualization, methodology, investigation, supervision, funding acquisition, and project administration.
2026/02/18
论文
arXiv
GeoAI
GIS
面向洪水淹没制图的新一代地理人工智能基础模型评估
Assessment of a new GeoAI foundation model for flood inundation mapping

视觉基础模型是地理空间人工智能(GeoAI)领域的前沿方向,该领域为地理空间问题求解与地理知识发现应用并拓展人工智能技术。由于其能够通过学习和提取海量地理空间数据中的重要图像特征,从而实现强大的图像分析能力,因此具有重要意义。本文评估了首个地理空间基础模型——IBM-NASA联合研发的Prithvi模型在关键地理空间分析任务——洪水淹没制图中的表现。将该模型与基于卷积神经网络及视觉Transformer的架构进行对比,评估其在洪水区域制图精度方面的性能。实验采用Sen1Floods11基准数据集,并基于测试数据集以及模型从未见过的全新数据集,评估各模型的预测能力、泛化能力与迁移能力。结果表明,Prithvi模型具备良好的迁移能力,尤其在未见过区域的洪水区域分割任务中表现出显著优势。研究同时指出,Prithvi模型在多尺度表征学习的采用、面向高层图像分析任务的端到端流程开发,以及输入数据波段灵活性方面仍有改进空间。

Wenwen Li, Hyunho Lee, Sizhe Wang
2023/09/26
论文
International Journal of Geographical Information Science
PublisherJournal
GeoAI
Min Deng Xiaoyong Tan Kaiqi Chen Baoju Liu Zhiyuan Zhao Youjun Tu Sheng Wu Xin Hu Zhiwei Zeng a Department of Geo-Informatics, School of Geosciences and Info-physics, Central South University, Changsha, Chinab The Third Surveying and Mapping Institute of Hunan Province, Hunan Geospatial Information Engineering and Technology Research Center, Changsha, Chinac Academy of Digital China (Fujian), Fuzhou University, Fuzhou, ChinaMin Deng is currently a doctoral supervisor and associate dean of the School of Geosciences and Info-Physics, Central South University, Changsha, China. His current research interests include coordinated planning, spatio-temporal data mining, spatio-temporal analysis, modeling and prediction. He has hosted numerous major projects, including a key project of the National Natural Science Foundation of China. He contributed to the conceptualization, methodology, formal analysis, validation, writing - original draft, writing - review & editing and funding acquisition.Xiaoyong Tan is currently a Ph.D. candidate in the School of Geosciences and Info-Physics, Central South University, Changsha, China. His research interests include spatio-temporal prediction and data mining. He contributed to the conceptualization, methodology, formal analysis, software, data curation, visualization, writing - original draft and writing - review & editing.Kaiqi Chen is currently a lecturer in the School of Geosciences and Info-Physics, Central South University, Changsha, China. His research interests include spatio-temporal prediction and data mining. He contributed to the conceptualization, methodology, resources, supervision, writing - original draft, writing - review & editing and funding acquisition.Baoju Liu is currently an associate professor and a doctoral supervisor in the School of Geosciences and Info-Physics, Central South University, Changsha, China. His research interests include spatio-temporal data mining and Geospatial optimization. He contributed to data curation, validation, visualization and funding acquisition.Zhiyuan Zhao is an Associate Researcher at Fuzhou University and the Deputy Director of the Digital Fujian Government Affairs Big Data Research Institute. His main research areas include population dynamics observation and application modeling, as well as regional digital development consulting. He contributed to data curation, formal analysis and validation.Youjun Tu is currently a Ph.D. candidate in Computer Science and Technology at Fuzhou University. His research interests include spatio-temporal big data mining, crowd dynamics observation, and application modeling. He contributed to data curation, formal analysis and validation.Sheng Wu is a Professor at Fuzhou University, Deputy Director of the Theoretical and Methodology Committee of the China Geographic Information Industry Association, and a member of the Cartography and Geographic Information System Professional Committee of the China Society for Geodesy, Photogrammetry, and Cartography. His main research areas include big data analysis and visualization, digital planning, and digital government. He contributed to data curation and funding acquisition.Xin Hu is currently a Ph.D. candidate in the School of Geosciences and Info-Physics, Central South University, Changsha, China. His research interests include spatio-temporal knowledge extraction and knowledge graph construction. He contributed to the data curation and formal analysis.Zhiwei Zeng is currently a master candidate in the School of Geosciences and Info-Physics, Central South University, Changsha, China. His research interests include spatio-temporal data mining. He contributed to the data curation and formal analysis.
2025/08/04
项目
GitHub Repositories
GeoAI
geographic intelligent agents
juaquicar/GeoAgents:面向地理空间AI智能体的开源框架,支持规划、多工具执行、假设验证、重新规划及全流程可追溯性
juaquicar/GeoAgents: Open-source framework for geospatial AI agents with planning, multi-tool execution, hypothesis verification, replan, and full traceability.

面向地理空间AI智能体(GeoAI)的开源框架,支持任务规划、多工具协同执行、假设验证、动态重新规划以及全流程操作可追溯性。涉及主题包括:ai-agents、django、geoai、geospatial、gis、llm。本GitHub仓库由开发者juaquicar维护,主要编程语言为Python,最后更新时间为2026-03-26。

juaquicar
2026/03/26
工具
GitHub Repositories
GeoAI
geospatial-analysis
GeoRetina/Arion:Arion 是由 GeoRetina Inc. 开发的首个面向地理空间分析的智能体式 AI 桌面应用
GeoRetina/Arion: Arion is the first agentic AI desktop app for geospatial analysis, developed by GeoRetina Inc.

Arion 是由 GeoRetina Inc. 开发的首个面向地理空间分析的智能体式 AI 桌面应用。相关主题包括:AI、AI 智能体(AI-agents)、地理空间 AI(GeoAI)、地理空间分析(geospatial-analysis)、地理信息系统(GIS)。本 GitHub 仓库由组织 GeoRetina 维护。主要编程语言为 TypeScript。GitHub 星标数:65。最后更新时间:2026-03-26。# Arion:面向智能体式地理空间 AI 分析的跨平台桌面应用 🎉 v0.10.1 已发布!Arion v0.10.1 现已上线,新增 Codex 集成、连接器标签页及用户界面优化。详见发布说明!📦 二进制版本即将发布!适用于 Windows、macOS 和 Linux 的二进制构建版本即将上线,敬请期待!Arion 是一款**跨平台桌面应用**,专为高级地理空间分析与智能体式工作流设计。基于 Electron、React(TypeScript)与 Vite 构建,Arion 原生支持 **Windows、macOS 和 Linux**,使用户能够利用本地及

GeoRetina
2026/03/26
论文
基于手机数据的情境感知人口位移估计:一种方法论框架
PaperDiscovery
Earth AI:利用基础模型与跨模态推理解锁地理空间洞察
Earth AI: Unlocking Geospatial Insights with Foundation Models and Cross-Modal Reasoning

数十年来,地理空间数据长期处于孤立状态……进而实现跨领域分析。为应对这一挑战,地理空间人工智能(GeoAI)(Iyer 等,2025)已从专用模型发展为面向地球观测的通用基础模型(Zhu 等,2024)。

2025/10/21
论文
arXiv
GeoAI
GIS
OmniGeo:面向地理空间人工智能的多模态大语言模型
OmniGeo: Towards a Multimodal Large Language Models for Geospatial Artificial Intelligence

多模态大语言模型(LLM)的快速发展为人工智能开辟了新领域,实现了文本、图像及空间信息等多样化大规模数据类型的融合。本文探讨了多模态大语言模型(MLLM)在地理空间人工智能(GeoAI)中的潜力,该领域利用空间数据应对地理语义、健康地理学、城市地理学、城市感知以及遥感等领域的挑战。我们提出一种专用于地理空间应用的MLLM(OmniGeo),能够处理和分析异构数据源,包括卫星影像、地理空间元数据和文本描述。通过结合自然语言理解与空间推理的优势,本模型提升了指令遵循能力以及GeoAI系统的准确性。实验结果表明,该模型在多种地理空间任务上优于特定任务模型及现有LLM,在处理多模态特性的同时,于零样本地理空间任务中取得了具有竞争力的表现。代码将在论文发表后公开。

Long Yuan, Fengran Mo, Kaiyu Huang
2025/03/21
论文
arXiv
GeoAI
GIS
GeoAgentBench:面向空间分析的工具增强型智能体动态执行基准
GeoAgentBench: A Dynamic Execution Benchmark for Tool-Augmented Agents in Spatial Analysis

大型语言模型(LLM)与地理信息系统(GIS)的融合标志着自主空间分析范式的转变。然而,由于地理空间工作流具有复杂、多步骤的特性,对这类基于LLM的智能体进行评估仍具挑战性。现有基准主要依赖静态文本或代码匹配,忽视了动态运行时反馈以及空间输出的多模态特性。为弥补这一空白,我们提出GeoAgentBench(GABench),一个专为工具增强型GIS智能体设计的动态交互式评估基准。GABench提供了一个真实的执行沙箱,集成了117个原子级GIS工具,覆盖6个核心GIS领域中的53类典型空间分析任务。鉴于精确的参数配置是动态GIS环境中执行成功的主要决定因素,我们设计了参数执行准确率(Parameter Execution Accuracy, PEA)指标,采用“最终尝试对齐”(Last-Attempt Alignment)策略,量化隐式参数推断的保真度。作为补充,我们提出一种基于视觉-语言模型(Vision-Language Model, VLM)的验证方法,用于评估数据-空间准确性及制图风格符合度。此外,为应对因参数错配和运行时异常导致的频繁任务失败,我们开发了一种新型智能体架构——Plan-and-React,该架构通过解耦全局编排与逐步响应式执行,模拟专家级认知工作流。针对七种代表性LLM开展的大量实验表明,Plan-and-React范式显著优于传统框架,在多步推理与错误恢复方面尤其突出,实现了逻辑严谨性与执行鲁棒性的最优平衡。我们的研究揭示了当前能力边界,并确立了一个稳健的

Bo Yu, Cheng Yang, Dongyang Hou
2026/04/15
论文
ACM
PaperDiscovery
面向地理空间人工智能的基础模型(愿景论文)| 第30届国际地理信息系统进展会议论文集
Towards a foundation model for geospatial artificial intelligence (vision paper) | Proceedings of the 30th International Conference on Advances in Geographic Information Systems

本文探讨了构建面向地理空间人工智能(GeoAI)的多模态基础模型所蕴含的潜力与挑战。我们首先通过在两项地理空间语义任务上测试现有大语言模型(LLMs)(如GPT-2和GPT-3)的性能,验证该思路的优势。

入库时间:2026/03/26
1 / 135