基于改进快速点特征直方图和双重迭代的点云配准

Point Cloud Registration Based on Improved Fast Point Feature Histogram and Double Iteration

为解决迭代最近点算法配准效率低且精度依赖于点云良好初始位姿的问题,提出了一种基于改进快速点特征直方图和双重迭代的点云配准方法,该方法包括点云粗配准和精配准两个步骤。点云粗配准首先利用体素滤波对初始点云进行预处理,再用内部形状描述子算法提取初始点云的特征点并求法向量,最后利用快速点特征直方图提取特征向量,根据法向量夹角和特征向量各自的差异进行粗配准,进而给精配准提供良好的初始位姿。点云精配准利用自适应阈值双重迭代最近点算法得到最终结果。采用两组点云进行试验并与三种传统配准方法作比较。试验结果表明,所提配准方法在主观视觉与客观评价指标方面均优于其它三种对比方法。所提方法能够克服配准点云数量限制,具有较高配准效率和精度。

In order to solve the problems such as the low registration efficiency of the iterative nearest point algorithm and the accuracy depending on the good initial posture of the point cloud,this paper proposes a point cloud coarse registration method based on the internal shape descriptor and the improved fast point feature histogram.Firstly,voxel filtering is used to prepro-cess the initial point cloud,and then the internal shape description sub-algorithm is adopted to extract the feature points of the initial point cloud and find the normal vector corresponding to the feature point,and the fast point feature histogram is used to extract the feature vector of the feature point.According to the difference between the angle of the feature vector and the nor-mal vector,the feature points are coarsely registered,which provides a good initial posture for the fine registration of the itera-tion of the nearest point registration algorithm.Experimental results show that the point cloud registration algorithm proposed in this paper can provide a better initial posture for the fine registration of the nearest point algorithm.Compared with several traditional registration methods,the proposed method overcomes the limit of the number of point clouds of registration,and further improves the efficiency and accuracy of registration.