基于点云配准的曲面三维差异检测  

3D Difference Detection of Surface Based on Point Cloud Registration

作　　者：罗锦鸿 陈新度[1] 吴磊[1] LUO Jin-hong;CHEN Xin-du;WU Lei(College of Mechanical and Electrical Engineering,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学机电工程学院,广州510006

出　　处：《组合机床与自动化加工技术》2021年第1期96-100,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：广东省省级科技计划项目:广东省智能制造技术与装备企业重点实验室(2017B030302004);广州市科技计划项目:面向抛光打磨的机器人智能共享工厂核心技术与装备研发(201902010054)。

摘　　要：在工件打磨任务中,实测数据与标准模型数据之间的差异性比较是实现机器人自动化打磨的关键研究问题之一。针对如何快速精确地提取陶瓷素胚工件曲面的三维差异数据问题,设计了一种曲面配准流程。首先,针对点云的粗对齐步骤,采用体素网格滤波进行下采样,采用了Hough投票法进行粗配准,将粗配准步骤运行时间控制在0.3 s以内。其次,针对点云的精确配准步骤,利用了CUDA加速比较耗时的ICP配准算法,将精确配准耗时控制在0.5 s以内。最后,针对差异提取步骤,通过引入Octree数据结构实现了差异点云的快速提取。最终,将模型曲面点云与扫描曲面点云之间的RMS误差控制在0.5 mm内,总运行时间控制在1 s之内,实现了两个曲面的快速精确对齐以及差异提取。In the task of workpiece polishing,the comparison of differences between the measured data and the standard model data is one of the key research problems to realize automatic polishing of robot.In this paper,a process of surface registration is designed to solve the problem of how to quickly and accurately extract the 3D difference data of workpiece surface.Firstly,for the coarse alignment step of point cloud,the down sampling is carried out by voxel grid filtering and coarse registration is carried out by Hough voting method,and the running time of coarse registration step is controlled within 0.3 s.Secondly,for the precise registration step of point cloud,the ICP registration algorithm of CUDA acceleration comparison time consumption is used to control the precise registration time within 0.5 s.Finally,aiming at the difference extraction step,we introduce Octree Data structure to realize the fast extraction of the difference point cloud.Finally,the RMS error between the model surface point cloud and the scanning surface point cloud is controlled within 0.5 mm,and the total running time is controlled within 1 s,which realizes the fast and accurate alignment and difference extraction of the two surfaces.

关 键 词：三维差异检测 点云配准 CUDA 霍夫投票 OCTREE 

分 类 号：TH16[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]