改进的迭代最近点点云配准方法

Improved Iterative Nearest Point Point Cloud Alignment Method

针对传统的迭代最近点(ICP)点云配准算法存在收敛缓慢、配准时间长、重叠率过低导致的匹配错误等问题,提出了一种以分块提取特征点为核心、块状配准点云重叠率为约束的改进ICP配准算法。首先,计算点云的平均距离密度,在设定的数量阈值内对点云进行分块,并从分块后的点云中并行提取尺度不变特征变换(SIFT)特征点,采用快速点特征直方图(FPFH)进行特征描述;然后,利用采样一致性初始配准(SAC-IA)算法实现点云的匹配,同时以块间匹配率50%作为依据,提取点云的重叠区域;最后,基于匹配的特征点计算初始姿态,在此基础上利用重叠部分实现两块点云的精确配准。实验结果表明,重叠率较低的点云经分块及重叠区域提取后,可以大幅缩短运行时间,提高配准精度。

Aiming at the problems of slow convergence,long alignment time,and matching error due to low overlap rate in the traditional iterative nearest point(ICP)point cloud alignment algorithms,an improved ICP alignment algorithm based on chunked feature point extraction as the core and chunked alignment point cloud overlap rate as the constraint is proposed.First,the average distance density of the point cloud is calculated,the point cloud is chunked within the set number threshold,and the scale invariant feature transform(SIFT)feature points are extracted in parallel from the chunked point cloud,and the fast point feature histogram(FPFH)is used for feature description;then,the sampling consistency initial alignment(SAC-IA)algorithm is used to realize the matching of the point cloud,and the overlapping region of the point cloud is extracted based on the 50%inter block matching rate;finally,the initial attitude is calculated based on the matched feature points,and the overlapping part is used to achieve accurate alignment of the two point clouds.The experimental results show that the point cloud with low overlap rate after segmentation and overlapping region extraction can greatly shorten the running time and improve the registration accuracy.