Guaranteed Outlier Removal for Point Cloud Registration with Correspondences (1711.10209v1)

Abstract: An established approach for 3D point cloud registration is to estimate the registration function from 3D keypoint correspondences. Typically, a robust technique is required to conduct the estimation, since there are false correspondences or outliers. Current 3D keypoint techniques are much less accurate than their 2D counterparts, thus they tend to produce extremely high outlier rates. A large number of putative correspondences must thus be extracted to ensure that sufficient good correspondences are available. Both factors (high outlier rates, large data sizes) however cause existing robust techniques to require very high computational cost. In this paper, we present a novel preprocessing method called \emph{guaranteed outlier removal} for point cloud registration. Our method reduces the input to a smaller set, in a way that any rejected correspondence is guaranteed to not exist in the globally optimal solution. The reduction is performed using purely geometric operations which are deterministic and fast. Our method significantly reduces the population of outliers, such that further optimization can be performed quickly. Further, since only true outliers are removed, the globally optimal solution is preserved. On various synthetic and real data experiments, we demonstrate the effectiveness of our preprocessing method. Demo code is available as supplementary material.

• The paper introduces a deterministic method, Guaranteed Outlier Removal (GORE), that prunes correspondence outliers for optimal 3D registration.
• It decomposes transformations for both 3 DoF and 6 DoF, reducing computational cost by limiting data for expensive global optimization.
• Extensive experiments show GORE’s effectiveness across robotics, surveying, and augmented reality, ensuring rapid and accurate alignment.

Overview of Guaranteed Outlier Removal for Point Cloud Registration

This paper introduces a preprocessing technique for 3D point cloud registration, focusing on effectively handling high rates of outliers without compromising the globally optimal solution. The research targets the challenge of aligning two point clouds, considering the transformation function derived from 3D keypoint correspondences heavily contaminated by outliers due to the reduced accuracy of 3D matches compared to their 2D counterparts. Robust registration techniques incur substantial computational costs when faced with large data sizes and high outlier rates.

Methodology

The proposed method, Guaranteed Outlier Removal (GORE), leverages geometric properties to efficiently discard matches that are guaranteed not to be part of the globally optimal solution. This deterministic approach significantly reduces data size before commencing more computationally expensive optimization processes. GORE is applicable to two fundamental registration paradigms:

1. Three-Degree of Freedom (3 DoF) Rotation: The technique involves decomposing the rotation into two partial transformations, employing bounding operations to define feasibility conditions, and executing interval stabbing algorithms to define angular constraints that optimize inlier count.
2. Six-Degree of Freedom (6 DoF) Rigid Transformation: By reframing the problem in terms of relative shifts and rotations among point correspondences, GORE extends its pruning capability to more general transformations, using branch-and-bound techniques selectively on reduced correspondence sets post-pruning.

Results and Implications

Extensive tests on both synthetic and real datasets, including mining, remote sensing, and RGB-D data, underscore the method's efficiency in drastically decreasing outlier proportions and data sizes, thereby facilitating robust point cloud registration in more feasible timeframes. For scenarios with very high outlier contamination, GORE combined with globally optimal algorithms, like branch-and-bound, offers practical improvements, signifying its utility in real-world applications where fast and reliable alignment is crucial.

The implications of this work are substantial for computer vision applications in robotics, surveying, and augmented reality involving large-scale 3D registration tasks. The method ensures that preprocessing steps do not compromise the integrity of subsequent global optimization tasks, making it a viable approach for real-time data processing in environments with limited computational resources.

Future Directions

The paper sets the stage for future advancements that could involve more robust outlier removal algorithms applicable to even more complex and varied 3D data scenarios. The integration of GORE with advanced AI-driven correspondence estimation techniques may further refine its application scope. Additionally, exploring parallel implementations and adaptive pruning strategies tailored to the specific characteristics of datasets could push the boundaries of current preprocessing efficiency even further.