基于表面间三棱锥体积测度的点云配准

Point Cloud Registration Using Surface Three-pyramid Volume Measure

提出一种基于离散对应特征和表面间平均三棱锥体积测度的改进ICP算法,进行了多视激光点云配准。通过寻找两片点云重叠区域内的有效点与对应三角形对,并将有效点与对应三角形对所夹的空间三棱锥体积作为误差测度来指导激光点云的配准;将点与对应三角形的质心作为对应点对,估计出新的空间位置转换关系。实验结果表明:该改进算法具有较高的配准精度且收敛速度较快。

An improved ICP（iterative closest points） algorithm was researched, which was based on discrete corresponding characteristics and error estimation of mean volume of three pyramid shaped by point and corresponding triangle between two surfaces, and it was used to align multi--view laser point cloud. Namely the registration of laser point cloud was instructed by searching effective point and corresponding triangle pair between two point cloud area and taking the spatial three--pyramid volume shaped by these point and corresponding triangle as error measure; then the new spatial position translation relation was estimated by using the point and the centroid of corresponding triangle as corresponding point pairs. The experimental results indicate the align accuracy of the improved algorithm is higher and its speed is faster than that of traditional ICP registration algorithm.