一种基于几何特征由粗到细点云配准算法

针对点云配准算法对初始位置敏感且收敛速度慢的问题,提出一种基于几何特征由粗到细点云配准算法。在粗配准阶段,通过投影法提取源点云和目标点云各4个轮廓点,然后利用曲率特征和轮廓点之间的距离寻找稳健的特征点对,计算得到初始刚性变换参数;细配准阶段,计算点云法向量及法向量夹角,以法向量为特征进行特征匹配,然后使用法向量夹角来启发搜索,使迭代最近点(iterative closest points, ICP)算法快速收敛。实验结果表明,所提出的由粗到细的配准算法鲁棒性强,具有较高的精度和速度。

A Coarse to Fine Point Cloud Registration Algorithm based on Geometric Features

To solve the problem that the point cloud registration algorithm is sensitive to the initial position and the convergence speed is slow, a coarse-to-fine point cloud registration algorithm based on geometric features was proposed. In the coarse registration stage, four feature points of the source point cloud and the target point cloud were extracted by the projection method, then the curvature feature and the distance between the matching points was used to find a robust feature point pair, finally the initial stiffness transformation parameters were calculated; In the fine registration stage, the point cloud normal vector and the normal vector angle were calculated, and the normal vector was used to match feature. Then the normal vector angle was used to inspire the search to make the two point cloud converge quickly. The results show that the proposed coarse-to-fine registration algorithm is robust and has high precision and speed.