点云 FPFH 特征提取优化配准算法

摘要

三维点云配准在机器人环境感知与建模、逆向工程等领域有着广泛的应用前景。针对获得的不同视角下点云数据的配准问题，设计一种 FPFH 特征提取优化配准方法。在提取关键点的基础上优化了FPFH 特征描述子计算过程中法向量计算，根据测量点及其邻域点估算每个关键点和它邻域点的曲面法矢，使用 SAC-IA 算法获得点云初始坐标变换矩阵，最后使用 ICP 算法精确配准。设计了三种配准方案，实验结果表明，只计算关键点及其周围一定范围内点法向量配准方法具有配准速度快、精度高的特点。%The three-dimensional point cloud registration has wide application prospect in fields of environmental perception and modeling of robot,reverse engineering,etc. An optimal feature extraction method of FPFH feature is designed to solve registration problem under different view of point cloud data. Based on the extracted key points,normal vector calculation used for getting FPFH feature descriptor is optimized. Based on the measurement point and its neighborhood points,the surface normal of each key points and its neighboring points are estimated. SAC-IA algorithm is used to obtain the initial point cloud coordinate transform matrix. Finally the ICP algorithm is used for precise registration. Three registration schemes are designed. The experimental results show that the scheme which only calculates normal of key points and their neighbors is faster and has high precision.