A three-dimensional scene procedural generation method and system based on semantic field driving

By using a semantic field-driven 3D scene generation method, the problem of mismatch between user intent and generated results is solved, enabling accurate generation and editing of 3D scenes, improving convenience, and making it suitable for various 3D scene generation needs.

CN122244402APending Publication Date: 2026-06-19HUNAN SANYUE SUWEI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUNAN SANYUE SUWEI TECH CO LTD
Filing Date
2026-05-22
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing 3D scene generation methods struggle to accurately understand complex user descriptions, resulting in generated results that do not match user intent. They also fail to maintain balance while satisfying various spatial constraints, impacting the scene's practicality and aesthetics.

Method used

A semantic field-driven approach is adopted. By receiving natural language instructions, semantic constraints are extracted, a semantic field function is constructed, and a weighted fusion and weight adjustment mechanism is used to optimize the comprehensive constraint field. The object position and rotation parameters are iteratively updated by combining a global energy function and a gradient descent algorithm to construct an invertible mapping table and generate an interactive 3D scene layout.

Benefits of technology

It achieves accurate conversion of natural language commands into three-dimensional spatial constraints, improves the accuracy of constraint field optimization, accelerates energy convergence speed, enhances the rationality of parameter configuration, and strengthens the convenience of scene editing, thereby reducing generation costs and production thresholds and adapting to various three-dimensional scene requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244402A_ABST
    Figure CN122244402A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for procedural generation of 3D scenes based on semantic field-driven approaches, belonging to the field of 3D scene generation technology. By extracting semantic constraints from instructions, a semantic field function is constructed to quantify the satisfaction distribution of objects in space. A weighted fusion and weight adjustment mechanism optimizes the comprehensive constraint field, ensuring a balance in the overall impact of multiple constraints. Global energy functions and gradient descent algorithms are used to evaluate parameter stability, and a swarm intelligence optimization algorithm is combined to iteratively update object position and rotation parameters to obtain the optimal geometric configuration. Simultaneously, a reversible mapping table is constructed to associate geometry with semantic constraints, supporting bidirectional queries and backtracking of the semantic field that contributes the most. Finally, an interactive 3D scene layout that allows real-time user modification is generated. This invention achieves intelligent transformation from instructions to scenes, improving the automation and interactivity of layout design and providing an efficient tool for constructing complex scenes.
Need to check novelty before this filing date? Find Prior Art