基于子空间特征向量的三维点云相似性分析

Similarity analysis of three-dimensional point cloud based on eigenvector of subspace

提出一种基于子空间特征向量的三维点云相似性分析算法。首先,获取两个物体的三维点云数据,并进行位置标准化。其次,利用最小子空间分割算法将两个三维点云分别分割成若干子空间。随后,计算子空间的质心到其拟合曲面的距离和夹角,并基于上述距离和夹角构成的向量空间,提取子空间特征向量。最后,通过特征向量间的相似度计算来评价两个三维点云的相似性。由于该方法将描述三维形体特征的子空间特征向量作为相似度度量的依据,所以具有数据量小、精度高的特点。实验表明,该算法能够定量地分析两个三维物体的相似性。

This paper presents a method of similarity analysis algorithm of the three-dimensional point cloud,which is based on eigenvector of the subspace. First of all, the three-dimensional point cloud data of two objects were obtained and positions of them were standardized. And then, the two three-dimensional point clouds were divided into several subspace by using the minimal spatial segmentation algorithm. Thirdly, the eigenvector of subspace were calculated, which should be divided into two steps：the first step was to calculate distance and angle from the centroid to the subspace surface, the next step was to compute the new eigenvector on the basis of vector space, which was composed of the distance and angle in step one. This research method took the advantage of small data in quantity and high precision in calculation because the eigenvector of subspace, which can describe the three-dimensional characteristics as the basis of similarity measure. The experiment shows that the algorithm can quantitatively analyze the similarity of two three-dimensional objects.