一种高度适用大模型的空间智能数据表达设计和定义方式

By reconstructing the physical world into discrete, atomized spatial entities and defining probabilistic features and semantic primitives, the conflict between determinism and probability in generating interactive digital earths from large models is resolved. Logical stitching and topological linking are achieved, improving the generation efficiency and accuracy of digital earths.

CN121811246BActive Publication Date: 2026-07-17YUNTU ZHIXING (BEIJING) TECHNOLOGY CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
YUNTU ZHIXING (BEIJING) TECHNOLOGY CO LTD
Filing Date
2025-12-25
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing technologies for generating interactive digital earths using large models suffer from problems such as conflicts between deterministic structures and probabilistic reasoning, lack of cross-grid semantic logic stitching, conflicts in multimodal observation data attributes, and a lack of unified data containers.

Method used

By employing a probabilistic extrapolation mechanism of generative large models, the physical world is reconstructed into discrete, atomized spatial entities. Probabilistic features and semantic primitives are defined, and the above problems are solved through normalized grid segmentation, logical stitching, and topological linking. A logical connection and fusion mechanism for cross-domain entities is constructed to achieve parallel generation and logical continuity of global-scale spatial data.

Benefits of technology

It achieves uncertainty compatibility in the large model generation process, ensures the logical continuity and accuracy of spatial data, supports the efficient fusion and dynamic evolution of multimodal observation data, and improves the generation efficiency and visualization quality of Digital Earth.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121811246B_ABST
    Figure CN121811246B_ABST
Patent Text Reader

Abstract

本申请涉及计算机数据处理与人工智能技术领域,尤其涉及一种高度适用大模型的空间智能数据表达设计和定义方式,包括:接收多模态观测数据作为输入;基于生成式大模型的概率推演机制;为每个所述原子化空间实体定义概率化特征;采用语义原语构建所述原子化空间实体的几何形态;生成包含所述概率化特征与语义原语的空间实体对象化表达模型。本申请摒弃了高词元消耗的网格面片,转而采用语义原语构建几何形态,使得空间数据能够以极低的词元成本被大模型理解、生成与存储。
Need to check novelty before this filing date? Find Prior Art