A driving scene understanding method based on spatial evidence constraint

By constructing spatial evidence units and reliability verification mechanisms, the problem of lack of spatial basis in the understanding results of driving scenarios in existing technologies has been solved, and reliable output in complex environments has been achieved.

CN122200604BActive Publication Date: 2026-07-24TONGJI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
TONGJI UNIV
Filing Date
2026-05-18
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing driving scenario understanding technologies lack spatial evidence constraints in complex environments, making it difficult to refute and verify the output results. Furthermore, they are prone to outputting unfounded conclusions when there is occlusion, perception degradation, or cross-modal conflict.

Method used

By constructing spatial evidence units, based on the time synchronization, spatial registration, and coordinate alignment of multimodal sensing data, the consistency residuals and uncertainty values ​​are determined, the mapping relationship between spatial evidence units and original sensing data is established, and reliability verification is performed to output normal, downgraded, or rejected results.

Benefits of technology

It achieves an explicit correspondence between the driving scenario understanding results and spatial basis, improves the interpretability and verifiability of the results, and ensures the reliability and robustness of the output in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122200604B_ABST
    Figure CN122200604B_ABST
Patent Text Reader

Abstract

The application discloses a driving scene understanding method based on spatial evidence constraint, and is applied to the technical field of automatic driving environment perception and scene understanding. The method comprises the following steps: acquiring multi-modal perception data of the environment around the vehicle; constructing an occupancy representation of a target scene to obtain occupancy states corresponding to a plurality of spatial units; determining consistency residuals and uncertainty values of the spatial units; establishing a mapping relationship between spatial evidence units and original perception data, and screening spatial evidence sequences; generating a candidate structured scene understanding result based on a scene understanding model; and performing reliability verification on the candidate structured scene understanding result to determine an output mode. The application can output a structured result containing a scene understanding conclusion, corresponding spatial evidence identification, a confidence value and an output mode; meanwhile, the reliability of the structured result is verified, so that the explainability, credibility and verifiability of the scene understanding result in a complex driving scene are improved.
Need to check novelty before this filing date? Find Prior Art