Multi-view point cloud registration method and device based on adaptive symmetry and hybrid optimization
By employing an adaptive symmetric and hybrid optimization multi-view point cloud registration method, an undirected topological graph is constructed using the YOHO network and spectral graph synchronization algorithm. Combined with Lie algebra space fine-tuning, the accuracy and stability issues of point cloud registration in complex scenarios are resolved, achieving high-precision global point cloud stitching.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUAQIAO UNIVERSITY
- Filing Date
- 2026-06-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multi-view point cloud registration methods are prone to generating a large number of serious mismatches in front-end feature matching under scenarios with weak features, high symmetry, and extremely low overlap rate. This leads to severe topological noise interference in back-end pose graph optimization, making it difficult to meet the needs of high-precision spatial mapping and digital twin construction in large-scale scenes.
An adaptive symmetric and hybrid optimization approach is adopted. Initial matching is performed through the YOHO 3D local feature extraction network. An undirected topological graph is constructed by combining dynamic edge number upper and lower bound constraints and logical union mechanism. The global solution is performed using the spectral graph synchronization algorithm and nonlinear fine-tuning is performed in the Lie algebra space to output high-precision global absolute pose.
It effectively solves the problems of noise interference with extremely low overlap rate in complex large scenes and the easy trap of local optimum initial value in traditional graph optimization, improves registration accuracy and stability, and realizes sub-millimeter level global point cloud stitching.
Smart Images
Figure CN122415697A_ABST