时空智能学:理论与应用

Spatiotemporal Intelligence:Theory and Applications

  • 摘要: 时空智能以统一时空基准为底座,依托通导遥算智能传感器、云计算、人工智能和领域知识,对运动的物质世界中的自然活动和人类活动进行自动、实时的感知与认知,并将结果转化为决策与行动支持。它既是测绘遥感地理信息学在智能时代的发展方向,也是人工智能进入运动的物质世界的重要组成部分。其核心任务是回答何时、何地、何种目标、何种变化、变化机理和何种决策的6W问题;在服务端则遵循4R要求,将经过核验的数据、信息与知识在任务规定时限内推送至需求发生地,交付实际使用者。分析了时空智能与当前以互联网数据为主的人工智能、大语言模型和遥感基础模型的关系,指出人工智能必须从互联网进一步走向物联网和运动的物质世界,形成感知智能、认知智能与行为智能相统一的闭环;综述了遥感自动定位与空中三角测量、目标识别、变化检测、分类以及从二维识别向四维认知演进的技术进展,介绍其在农业、交通、矿山、生态保护、公共安全、健康和社会经济等领域的应用。时空智能要坚持任务牵引、知识引导、独立自主、安全可用,使机器承担人类不能做、不愿做和危险繁重的劳动,降低劳动强度,提高劳动效率,服务联合国可持续发展目标和人类命运共同体建设,让人更聪明、让生活更美好,促进人与自然协同可持续发展。

     

    Abstract: Built on a unified spatiotemporal reference framework, spatiotemporal intelligence (STI) integrates communication, navigation, remote sensing, computing, artificial intelligence (AI), cloud computing, and domain knowledge to enable automatic and real-time perception and understanding of natural and human activities in the dynamic physical world, and to translate such understanding into support for decision-making and action. STI represents both an important direction for the development of surveying, remote sensing, and geographic information science in the intelligent era and a key pathway for extending AI from the digital domain to the physical world. Its core task is to address six fundamental questions, namely when, where, what object, what change, why the change occurs, and what decision or action should be taken. At the service level, STI follows the 4R principles by delivering right data, information, and knowledge to the right person, at the right place, and within the right time. We discuss the relationships between STI and Internet-data-centric AI, large language models, and remote sensing foundation models, and argue that AI needs to further extend from the Internet toward the Internet of things and the dynamic physical world, forming a closed loop that integrates perceptual, cognitive, and behavioral intelligence. Progress in automatic geolocation and aerial triangulation, object recognition, change detection, classification, and the transition from two-dimensional recognition to four-dimensional spatiotemporal understanding is reviewed, together with representative applications in agriculture, transportation, mining, ecological conservation, public safety, human health, and socio-economic development. The development of STI should remain task-driven, knowledge-guided, independently controllable, secure, and practically usable, enabling machines to undertake work that humans cannot perform, are unwilling to perform, or that is dangerous, strenuous, and repetitive. STI can reduce labor intensity, improve productivity, support the United Nations sustainable development goals and the building of a community with a shared future for humanity, improve quality of life, and promote the coordinated and sustainable development of humanity and nature.

     

/

返回文章
返回