Hierarchical Optimization of 3D Point Cloud Registration

Huikai Liu,Yue Zhang,Linjian Lei,Hui Xie,Yan Li,Shengli Sun

DOI: https://doi.org/10.3390/s20236999

IF: 3.9

2020-12-07

Sensors

Abstract:Rigid registration of 3D point clouds is the key technology in robotics and computer vision. Most commonly, the iterative closest point (ICP) and its variants are employed for this task. These methods assume that the closest point is the corresponding point and lead to sensitivity to the outlier and initial pose, while they have poor computational efficiency due to the closest point computation. Most implementations of the ICP algorithm attempt to deal with this issue by modifying correspondence or adding coarse registration. However, this leads to sacrificing the accuracy rate or adding the algorithm complexity. This paper proposes a hierarchical optimization approach that includes improved voxel filter and Multi-Scale Voxelized Generalized-ICP (MVGICP) for 3D point cloud registration. By combining traditional voxel sampling with point density, the outlier filtering and downsample are successfully realized. Through multi-scale iteration and avoiding closest point computation, MVGICP solves the local minimum problem and optimizes the operation efficiency. The experimental results demonstrate that the proposed algorithm is superior to the current algorithms in terms of outlier filtering and registration performance.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

This paper focuses on the problem of 3D point cloud registration, which is a key technology in the fields of robotics and computer vision. Currently, the most commonly used methods are Iterative Closest Point (ICP) and its variants, but these methods have issues such as sensitivity to outliers, large influence of initial pose, and low computational efficiency. The paper proposes a hierarchical optimization approach, including an improved voxel filter and Multi-Scale Voxelized Generalized ICP (MVGICP), to address these issues.

In terms of voxel filtering, the paper proposes a combination of traditional voxel sampling and point density information to effectively remove outliers and reduce the size of point clouds. Compared to traditional methods, the improved voxel filter improves computational efficiency while maintaining filtering performance.

In 3D point cloud registration, MVGICP utilizes a distribution-to-distribution strategy to avoid nearest point calculation, solving the problems of ICP getting easily trapped in local optima and low computational efficiency. Through multi-scale iterations, MVGICP achieves initial transformation without coarse registration and further refines the registration results, improving registration accuracy and efficiency.

Experimental results show that the proposed algorithm outperforms current algorithms in outlier filtering and registration performance. In summary, the main contributions of the paper include:

1. Proposing an improved voxel outlier filtering method that combines voxel sampling and point density, achieving accurate outlier filtering while improving downsampling efficiency.
2. Proposing a new 3D point cloud registration algorithm, MVGICP, which adopts multi-scale iterations, avoids complicated nearest point calculation, and significantly improves GICP's issue of local minima and computational efficiency.
3. Conducting comprehensive evaluation of the proposed algorithm, including comparison with other algorithms in outlier filtering and registration performance. The experiments demonstrate that the proposed method performs better than other algorithms on multiple datasets.

The subsequent sections of the paper provide detailed introductions to related work, methodology, experimental results and analysis, as well as conclusions.