基于FPFH-ICP的三维工件点云配准方法研究  

Research on 3D Artifact Point Cloud Registration Method Based on FPFH-ICP

作　　者：张泽鹏 赫森奥 张明哲 曹兵兵 张坤[1] ZHANG Zepeng;HE Senao;ZHANG Mingzhe;CAO Bingbing;ZHANG Kun(College of Mechanical and Electrical Engineering,Xi’an Polytechnic University,Xi’an Shaanxi 710600,China)

机构地区：[1]西安工程大学机电工程学院,陕西西安710600

出　　处：《信息与电脑》2022年第13期54-57,共4页Information & Computer

基　　金：陕西省大学生创新创业训练计划项目(项目编号:S202110709064)。

摘　　要：针对三维工件点云配准方法存在配准精度低等问题,笔者提出一种基于快速点特征直方图(Fast Point Feature Histogram,FPFH)初始匹配与迭代最近点(Iterative Closest Point,ICP)的点云配准方法。首先,在配准前使用随机抽样一致(Random Sample Consensus,RANSAC)算法对初始点云数据进行预处理,获得目标点云数据;其次,使用FPFH算法提取点云特征,进行特征匹配,求得初始变换矩阵;再次,使用ICP算法对初始变换矩阵进行增益,求得最终变换矩阵;最后,完成点云精确配准。基于三维工件点云,将本文算法与经典算法对比测试。实验结果表明,本文算法在配准精度上有所提高。Aiming at the problem of low registration accuracy of the 3D workpiece point cloud registration method,this paper proposes a point cloud registration method based on the initial matching and Iterative Closest Point(ICP)of the Fast Point Feature Histogram(FPFH).Firstly,the initial point cloud data is preprocessed by the Ransac algorithm before registration to obtain the target point cloud data.Secondly,the FPFH algorithm is used to extract the point cloud features,perform feature matching,and obtain the initial transformation matrix.Thirdly,the ICP algorithm is used to gain the initial transformation matrix to obtain the final transformation matrix.Finally,the point cloud is accurately registered.Based on the 3D workpiece point cloud,the comparison test and the classical algorithm show that the registration accuracy is improved.

关 键 词：三维工件 点云配准 配准精度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]