A camera pose visual measurement method based on three-dimensional gaussian splash

By using perturbation pose rendering and geometric constraint optimization based on a 3D Gaussian splash model, the applicability and efficiency of camera pose measurement methods in different scenarios are solved, achieving pose measurement with high accuracy and robustness.

CN122066779BActive Publication Date: 2026-06-26BEIHANG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
BEIHANG UNIV
Filing Date
2026-04-17
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing camera pose measurement methods have low applicability and deployment efficiency in different scenarios, and fail to fully consider the geometric relationships between new views, affecting the accuracy of pose measurement.

Method used

Based on the 3D Gaussian splash model, multiple perturbation poses are generated by applying perturbation to the initial pose. The 3D Gaussian splash model is then used to render the image for 2D-2D feature matching. The objective function, which combines geometric consistency and epipolar geometric constraints, is then iteratively optimized to determine the camera pose.

Benefits of technology

It improves the accuracy and efficiency of camera pose measurement in different scenarios, enhances adaptability and robustness, and eliminates the need for additional training or fine-tuning for different scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122066779B_ABST
    Figure CN122066779B_ABST
Patent Text Reader

Abstract

The application discloses a camera pose visual measurement method based on three-dimensional Gaussian splash, and relates to the field of camera pose measurement.The method comprises the following steps: acquiring an RGB image and a three-dimensional Gaussian splash model of a target scene and an initial pose of a camera; applying several disturbances to the initial pose to generate several disturbed poses; rendering the initial pose and each disturbed pose based on the three-dimensional Gaussian splash model to obtain L+1 rendered images; performing 2D-2D feature matching between the RGB image of the target scene and each rendered image; determining the relative poses between each rendered image and the RGB image of the target scene based on L+1 groups of matching points; taking the minimization of an objective function as a target to determine a pose update value, update the initial pose and return to the step of applying several disturbances to the initial pose; and repeating the above steps until a convergence condition is reached.The application can improve the accuracy and efficiency of pose measurement.
Need to check novelty before this filing date? Find Prior Art