| 摘 要: 迭代最近点(ICP)配准算法需要两点云处于良好的初始位置，否则在配准时容易陷入局部最优。针对该问题，提出了一种基于内部形状描述子(ISS)特征点与改进描述子的粗配准方法，使得低重叠度或无公共重叠部分的点云获取良好的初始位置。首先，利用ISS特征点提取点云的特征点；其次，基于特征点与其邻域点法向量夹角提出改进的描述子，根据描述子的欧氏距离将点云特征进行匹配；第三，通过单次最优变换进行粗配准；最后，对两点云进行精配准ICP迭代。实验表明，在点云模型完整的情况下，本文方法可为精配准提供良好的初始位置，且粗配准精度比传统点云配准精度高三个量级，配准效率提升23.7%。 |
| 关键词: 点云粗配准；ISS特征点；改进描述子；ICP算法 |
|
中图分类号: TP391.1
文献标识码: A
|
|
| Research on Point Cloud Registration Algorithm based on ISS Feature Points and Improved Descriptor |
|
XIA Kanqiang
|
(College of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China)
1121834925@qq.com
|
| Abstract: Iterative closest point registration algorithm (ICP) requires two point clouds to be in a good initial position, otherwise it is easy to fall into a local optimum during registration. In view of this problem, this paper proposes a coarse registration method based on Intrinsic Shape Signatures (ISS) feature points and improved descriptors, which makes point clouds with low overlap or without common overlap obtain a good initial position. Firstly, ISS feature points are used to extract the feature points of the point cloud. Secondly, an improved descriptor is proposed based on the angle between the feature points and the normal vector of the neighboring points, and the point cloud features are matched according to the Euclidean distance of the descriptor. Then, rough registration is carried out by single optimal transformation. Finally, precision registration ICP iteration is conducted for the two-point clouds. Experimental results show that the proposed method provide a good initial position for fine registration under the condition of complete point cloud model; the coarse registration accuracy is three orders of magnitude higher than that of traditional point cloud registration, and the registration efficiency is improved by 23.7%. |
| Keywords: point cloud coarse registration; ISS feature points; improved descriptor; ICP algorithm |
