Geometry Preserving Sampling Method based on Spectral Decomposition for 3D Registration
This addresses the computational bottleneck in 3D mapping for mobile robotics, offering an incremental improvement over prior sampling strategies.
The paper tackles the problem of reducing point cloud size for efficient 3D registration in mobile robotics, where existing methods underperform with non-uniform density and noise, by proposing a novel sampling algorithm called Spectral Decomposition Filter (SpDF) that preserves geometry and scales effectively.
In the context of 3D mapping, larger and larger point clouds are acquired with LIDAR sensors. The Iterative Closest Point (ICP) algorithm is used to align these point clouds. However, its complexity is directly dependent of the number of points to process. Several strategies exist to address this problem by reducing the number of points. However, they tend to underperform with non-uniform density, large sensor noise, spurious measurements, and large-scale point clouds, which is the case in mobile robotics. This paper presents a novel sampling algorithm for registration in ICP algorithm based on spectral decomposition analysis and called Spectral Decomposition Filter (SpDF). It preserves geometric information along the topology of point clouds and is able to scale to large environments with non-uniform density. The effectiveness of our method is validated and illustrated by quantitative and qualitative experiments on various environments.