Point Cloud Registration Algorithm based on Feature Histogram and Fast Anomaly Removal
To improve the accuracy and efficiency of point cloud registration,particularly for adjacent viewpoint point clouds that are heavily influenced by noise,have low feature extraction and matching efficiency,and suffer from point cloud redundancy,this paper presents a novel and efficient feature histogram point cloud registration algorithm(EFH-PCR).The EFH-PCR algorithm first reduces the number of points in the point cloud through voxel grid down-sampling,and then extracts key points by the 3D-Harris algorithm.Next,a target function based on feature distribution histograms is constructed for feature matching and robustness constraints.This target function is used to determine corresponding feature point pairs between point clouds and optimize the registration results.Finally,a global optimization algorithm is used to solve the target function,and its result is further optimized by the ICP algorithm.Experimental results show that the proposed algorithm outperforms Super 4PCS in terms of multi-view point cloud registration,with a 12％increase in overlap rate and a 43％reduction in computation time,demonstrating its advantages in fast registration.
3D point cloudpoint cloud registrationfast point feature histogramoutlier removalnearest point iteration algorithm
NETL
NSTL
万方数据
