空间非合作目标点云聚类配准方法

Improved registration algorithm for spatial non-cooperative target point cloud clustering

空间非合作目标的相对位姿测量问题成为空间在轨操作任务的重难点,通过对激光雷达获取的目标三维点云进行聚类,得到小规模、特征明显的聚类点云,有效提高了配准效率和精度。针对基于区域生长的聚类算法在对可视点云进行聚类时,特征相似部分无法聚类识别的问题,提出了二维图像优化三维点云聚类的方法。该方法将深度值信息和RGB颜色值建立数学映射关系,点云降维后,利用颜色梯度突变进行边界提取,将边界内的点逆向恢复到原始点云,最后将各个类的点云进行合并,得到易于识别的显著特征点云。实验结果表明,在配准角度误差为±5°的条件下,可有效地缩减点云规模并保留了显著特征,提高ICP配准算法的计算效率,为解决空间非合作目标相对位姿实时测量提供技术支持和解决思路。

The problem of relative pose measurement of non-cooperative targets in space becomes the most difficult task of the in-orbit operation task. By clustering the three-dimensional point cloud of the target obtained by Lidar, a small-scale cluster point cloud with obvious characteristics is obtained, which effectively improves the registration efficiency and precision. Aiming at the problem that clustering algorithm based on region growth can not recognize clusters with similar features when clustering visible point clouds, a method for optimizing 3D point cloud clustering with 2D images was proposed. This method established a mathematical mapping relationship between the depth value information and the RGB color value. After the point cloud dimension was reduced, the boundary was extracted using the color gradient mutation, the points within the boundary were reversely restored to the original point cloud, and finally the various types of point clouds were combined to obtain a point cloud with distinctive features that were easy to recognize. The experimental results show that under the condition of the registration angle error is ±5°, the size of the point cloud can be effectively reduced and the significant features are preserved. It provide technical support and solutions for the real-time measurement of relative pose of non cooperative targets in space.