A three-dimensional gauss space-ground cross-view vehicle positioning method and system based on vector line guidance
The 3D Gaussian air-to-ground cross-view vehicle localization method guided by vector lines solves the problems of insufficient robustness and accuracy of air-to-ground cross-view localization technology under complex visual conditions, achieves higher accuracy vehicle localization, reduces the influence of artifacts and texture blurring, and improves the geometric accuracy of pose calculation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN UNIV
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-26
AI Technical Summary
Existing air-to-ground cross-view positioning technologies lack robustness and accuracy under complex visual conditions. In particular, when dealing with differences in viewpoint, viewpoint distortion, scale changes, and illumination between aerial and ground images, traditional local feature descriptors struggle to establish stable matching relationships. Furthermore, artifacts, blurring, and texture stretching issues exist during 3D reconstruction, and ghosting or depth smoothing effects are prone to occur during depth map generation, affecting the geometric accuracy of pose calculation.
A vector-line-guided 3D Gaussian air-ground cross-view vehicle localization method is adopted. Vector lines are extracted through sparse reconstruction and dense point cloud generation to determine the pose of the ground camera and generate a virtual view. The image is repaired using a diffusion model and a depth map is generated. Combined with pose calculation technology, the localization accuracy and robustness are improved.
It improves the robustness and accuracy of cross-view positioning between air and ground under complex visual conditions, ensures the accuracy and stability of vehicle positioning, reduces the impact of artifacts and texture blurring, and improves the geometric accuracy of pose calculation.
Smart Images

Figure CN121883600B_ABST
Abstract
Citation Information
Patent Citations
Three-dimensional Gaussian visual positioning method for sparse visual angle
CN120339380A
Camera pose estimation method based on 2D Gaussian splashing
CN121661143A