从大模型看测绘时空信息智能处理的机遇和挑战  

Opportunities and Challenges of Spatiotemporal Information Intelligent Processing of Surveying and Mapping in the Era of Large Models

作　　者：杨必胜[1] 陈一平 邹勤[3] YANG Bisheng;CHEN Yiping;ZOU Qin(State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan 430079,China;School of Geospatial Engineering and Science,Sun Yat-Sen University,Zhuhai 519082,China;School of Computer Science,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]中山大学测绘科学与技术学院,广东珠海519082 [3]武汉大学计算机学院,湖北武汉430072

出　　处：《武汉大学学报（信息科学版）》2023年第11期1756-1768,共13页Geomatics and Information Science of Wuhan University

基　　金：国家自然科学基金(42130105,62171324);国家重点研发计划(2022YFB3904100)。

摘　　要：当前,时空信息、定位导航服务已成为重要的新型基础设施,在通用人工智能的驱动下,大模型引领的智能时代已经到来,越来越强大的大模型将在测绘时空信息智能处理与应用中发挥越来越重要的作用。以大模型为研究范式,总结大模型在测绘时空信息智能处理的现状与进展,分析大模型在测绘时空信息智能处理面临的挑战,阐述多模态融合与理解架构设计、提示工程优化微调以及人在回路引导决策3个核心方面在时空信息测绘大模型中的关键作用。针对行业理解深度、数据安全隐患、内容可信度保障以及训练部署成本优化4个方面,展望时空信息测绘大模型面临的挑战与发展趋势。Currently,spatiotemporal information,positioning and navigation have become important new infrastructures.Driven by general artificial intelligence,the era of intelligence led by large models has arrived.The increasingly powerful large models will play a more and more important role in the intelligent processing and applications of spatiotemporal information.First,this paper uses large models as the research paradigm and summarizes the current status and progress of large models in intelligent processing of spatiotemporal information in surveying and mapping.In addition,it analyzes the challenges faced by large models in intelligent processing of spatiotemporal information in surveying and mapping.We elaborate three key technologies in spatiotemporal information large models for surveying and mapping,including multi-modal fusion and understanding architecture design,prompt engineering optimization by fine-tuning,and human-in-theloop guidance for decision-making.Finally,the depth of industry understanding,data security risks,content credibility,and training deployment cost optimization are addressed to predict the development trend of spatiotemporal information large models.

关 键 词：时空信息 大模型 多模态 提示工程 人在回路 智能化测绘 

分 类 号：P208[天文地球—地图制图学与地理信息工程]