基于重叠域采样混合特征的点云配准算法

Point cloud registration algorithm based on mixed-features sampling for overlapping domain

针对部分重叠的两片点云配准效率低、误差大等问题,提出了一种基于重叠域采样混合特征的点云配准算法。首先,通过编码和特征交互的方式预测每个点的重叠分数,获得更丰富的点云特征;其次,提取重叠点的局部几何特征,基于重叠分数和点特征的显著性保留重叠关键点;最后,利用重叠关键点的几何信息和空间信息构建混合特征矩阵,计算矩阵的匹配相似度,采取加权奇异值分解运算得到配准结果。实验结果表明,该方法具有较强的泛化能力,能在保证配准效率的同时显著提升点云配准精度。

Aiming at the problems of low efficiency and large error in the registration of two partially overlapped point clouds,this paper proposed a point cloud registration algorithm based on the mixed-features sampling for overlapping domain.Firstly,it used coding and feature interaction predicted the overlap score of each point to obtain richer point cloud features.Secondly,it extracted the local geometric features of overlapping points and retained the overlapping key points based on overlapping scores and the significance of point features.Finally,it used the geometric information and spatial information of overlapping key points to construct a hybrid feature matrix,calculated the matching similarity of the matrix,and obtained registration results by using weighted singular value decomposition.Experimental results show that the proposed method has strong generalization ability and can significantly improve the registration accuracy of point cloud while ensuring registration efficiency.