Abstract:

The registration between point cloud data with low overlap rate and large angle is studied. A global point cloud initial registration algorithm based on fractal dimension is proposed. Firstly, the dimension value of each point in the point cloud is calculated, the feature points are extracted from the point cloud by the dimension attribute, the feature points are clustered to form a global structure, and from the global structure, the congruent triangle pair is obtained as the matching point pairs for initial registration, finally the Trimmed Iterative Closest Point（Trimmed-ICP） fine registration algorithm is performed. Compared with the Global optimal Iterative Closest Point（Go-ICP） algorithm, the proposed algorithm can effectively reduce the pose difference between point cloud data at different angles, and has a significant effect on the registration of point cloud data with low overlap rate and large angle.

Key words: global optimal iterative closest point algorithm, fractal dimension, global structure, point cloud initial registration

摘要：

针对重叠率低、角度大的点云数据之间的配准进行了研究，提出基于分形维数的全局点云初始配准算法。计算点云中各点的维数值；通过维数属性，从点云中提取特征点；聚类特征点，形成全局结构；从全局结构中，获得全等三角形对，作为匹配点对，进行初始配准；进行剪枝迭代最近点（Trimmed Iterative Closest Point，Trimmed-ICP）细配准。该算法与全局最优迭代最近点（Global optimal Iterative Closest Point，Go-ICP）算法相比，能够有效缩小不同角度的点云数据之间的位姿差异，显著提升对重叠率低、角度大的点云数据的配准效果。

关键词: 全局最优迭代最近点算法, 分形维数, 全局结构, 点云初始配准