.png){kind=icon}
.png){kind=icon}
Return
GCG-PROSAC: a geometric consistency-based robust estimation algorithm for accelerated 3D point cloud registration
.png){kind=icon}
PRE
AI
Z
X
DOI：10.1016/j.displa.2026.103430.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Three-dimensional (3D) point cloud registration plays a vital role in applications including virtual and augmented reality, computer graphics, and human-computer interaction. However, due to environmental occlusions, sensor field-of-view limitations, and constraints in feature extraction, a large number of outliers often exist in the established correspondences, making it difficult to rapidly establish a robust and stable rigid transformation between point clouds. This paper proposes a robust estimation algorithm, Geometric Consistency Graph-Progressive Sample Consensus (GCG-PROSAC), for fast computation of rigid transformations. The algorithm first utilizes geometric consistency on the point cloud surface during local optimization to construct a new geometric consistency graph, which improves the accuracy and speed of parameter solving in each iteration. Additionally, based on geometric consistency, the algorithm introduces a novel matching quality-based sampling strategy to accelerate the process of finding the optimal rigid transformation. Compared to existing robust estimation algorithms, GCG-PROSAC demonstrates higher accuracy across various scenarios. Compared to the baseline algorithm, under Gaussian noise, GCG-PROSAC reduces angular and translational errors by 55.27% and 65.12%, respectively. In occluded environments, these reductions reach 95.98% and 97.30%, significantly accelerating convergence. This highlights GCG-PROSAC's superior performance and faster convergence rate over traditional methods in challenging conditions.
Keywords:
Point cloud registration
Robust estimation
RANSAC variants
Geometric consistency
Journal
IF:
3.4
Papers:
2.1K
Citations:
3.2K
Organization
Cited Papers