针对超四点快速鲁棒匹配算法(Super-4PCS)粗匹配过程计算复杂度较高,配准时间长等问题,提出一种结合改进快速点特征直方图(FPFH)的Super-4PCS粗配准算法。通过主成分分析法(PCA)从快速点特征直方图中筛选出能代表点云特征信息的特征点,并将筛选出的特征点云作为输入数据进行Super-4PCS粗配准,由Super-4PCS粗配准得到初始变换矩阵,再进一步进行最近点迭代算法(ICP)精配准。为了验证在不同密度点云下的匹配效率,分别使用Bunny、Dragon两种不同密度的点云数据集进行配准实验,在满足精配准精度的基础上,对比FPFH-SAC和Super-4PCS粗配准方法,粗配准速率分别提升了72%和58%,总体配准速率分别提升了43%和32%。

In response to the problems of high computational complexity and long alignment time in the coarse matching process of the Super-4 Points Fast Robust Matching Algorithm (Super-4PCS), a Super-4PCS coarse alignment algorithm combined with the improved Fast Point Feature Histogram (FPFH) is proposed. The feature points that can represent the feature information of the point cloud are filtered from the fast point feature histogram by principal component analysis (PCA), and the filtered feature point cloud is used as the input data for Super-4PCS coarse alignment, and the initial transformation matrix is obtained from Super-4PCS coarse alignment, and then the nearest point iteration algorithm (ICP) fine alignment is further performed. Two points cloud datasets with different densities, Bunny and Dragon, are used for the alignment experiments to verify the matching efficiency under different densities of point clouds. Based on satisfying the fine alignment accuracy, compared with the FPFH-SAC and Super-4PCS coarse alignment methods, the coarse alignment rate is improved by 72% and 58% respectively, and the overall alignment rate is improved by 43% and 32% respectively.

TN957. 52

国家自然科学基金资助项目(61901400);四川省自然科学基金资助项目(2022NSFSC0542)