一种低重叠率激光点云的配准方法

A registration method of laser point cloud with low overlap

针对点云数据采集过程中因扫描仪设站数少导致相邻的两片激光点云重叠率低,且难以高精度进行配准的问题,该文提出了一种基于重叠区域的点云配准方法。首先利用加入距离权重的法线夹角及曲率特征将点云分割成块,构建每一点云块的多维特征描述符,通过比较各点云块间的方差分布相似性提取相邻点云的重叠区域,然后将重叠区域的点云带入超四点快速鲁棒匹配(Super4PCS)算法中进行配准,根据一致性约束将最优的刚性变换矩阵应用于原始数据,得到最终的点云配准模型,最后与直接利用Super4PCS算法配准后的效果进行对比分析。实验结果表明,通过增加点云间重叠区域的提取,可以有效提高对低重叠率激光点云的配准精度,从而更有利于点云三维模型构建等后续数据处理。

Aiming at the problem that the overlap rate of two adjacent laser point clouds is low due to the small number of scanners in the point cloud data acquisition process,and it is difficult to register with high precision,this paper proposed a point cloud registration method based on overlapping regions.Firstly,the point cloud was divided into blocks according to the angle of the normal and the curvature added the distance weights,the multi-dimensional feature descriptor of each point cloud block was constructed.The overlap of adjacent point clouds was extracted by comparing the similarity of variance distribution between each point cloud block.Then,taking the point cloud of the overlap region as input data,matching is performed by using the super 4-points congruent sets(Super4 PCS)algorithm,and the optimal rigid transformation matrix was applied to the original data according to the consistency constraint to obtain the final point cloud registration model.Finally,the registration results of using Super4 PCS algorithm directly were compared and analyzed.The experimental results showed that by increasing the extraction of overlapping regions between point clouds,the registration accuracy of low overlap laser point clouds could be effectively improved,which is more conducive to subsequent data processing such as point cloud 3 Dmodel construction.