摘要
针对传统2D视觉下机械臂无法抓取具有3D位姿物体的问题,提出了基于3D视觉的机械臂抓取控制方法。该方法首先通过差分和滤波的方式去除噪点,并提出了一种加权下采样的方法简化点云。通过欧式距离和法线夹角变化率的结合,提出了一种点云二次分割方法。物体的定位方式采用基于主成分分析法(PCA)的粗配准算法和基于迭代最近点算法(ICP)的精配准算法,然后将所得到的目标物体位姿发送给机械臂进行抓取。在达明协作机器人上的实验表明,该方法能够有效抓取具有3D位姿的物体,与已有的方法相比,在精度和处理时间上有了较大的提升。
Robotic arms under traditional 2D vision cannot grasp objects with 3D pose.Aiming at the problem,this paper proposes a robotic grasping control method based on 3D vision.The noise is removed by difference and filtering at first,and a weighted down sampling method is proposed to simplify the point cloud in this work.By using the combination of Euclidean distance and normal angle change rate,a secondary segmentation method of point cloud is proposed.The positioning method of the object adopts the coarse registration algorithm based on the principal component analysis(PCA)algorithm and the fine registration algorithm based on the iterative closest point(ICP)algorithm,and then sends the obtained pose of the target object to the robotic arm for grasping.Experiments on TM collaborative robot show that the proposed method can grasp objects with 3D poses effectively.Compared with the existing methods,the accuracy and processing time of the proposed method are greatly improved.
作者
潘聪
方灶军
张群莉
孙晨阳
戴伊宁
PAN Cong;FANG Zaojun;ZHANG Qunli;SUN Chenyang;DAI Yining(College of Mechanical Engineering,Zhejiang University of Technology,Hangzhou 310014;Ningbo Institute of Materials Technology and Engineering,Chinese Academy of Sciences,Ningbo 315201;Zhejiang Key Laboratory of Robot and Intelligent Manufacturing Equipment Technology,Ningbo 315201)
出处
《高技术通讯》
CAS
2023年第12期1313-1322,共10页
Chinese High Technology Letters
基金
国家自然科学基金(U1909215,92048201,52127803)
浙江省重点研发计划(2019C01043,LD22E050007)
中科院装备研制项目(YJKYYQ20200030)
宁波市2025重大专项(2021Z020)资助项目。
关键词
3D视觉
机械臂抓取
点云处理
位姿估计
定位精度
3D vision
robot grabbing
point cloud processing
pose estimation
positioning accuracy
