基于深度图像位姿识别和机构运动学的自动抓取管片试验

庄欠伟; 朱雁飞; 于宁; 黄德中; 杨正; 马志刚; 袁玮皓

doi:10.3973/j.issn.2096-4498.2024.11.003

隧道建设 >

2024 , Vol. 44 >Issue 11: 2133 - 2138

DOI: https://doi.org/10.3973/j.issn.2096-4498.2024.11.003

研究与探索

基于深度图像位姿识别和机构运动学的自动抓取管片试验

庄欠伟 ,
朱雁飞 ,
于宁 ,
黄德中 ,
杨正 ,
马志刚 ,
袁玮皓

展开

（1. 上海隧道工程有限公司，上海 200032； 2. 上海城建隧道装备有限公司，上海 200137； 3. 上海申通地铁集团有限公司，上海 200122）

庄欠伟（1978—），男，山东临沂人，2005年毕业于浙江大学，机械电子专业，博士，正高级工程师，现从事隧道机械技术研究与试验工作。E-mail： qianweizhuang@sohu.com。

网络出版日期: 2024-12-12

收起

Experimental Study of Automatic Segment Grasping Based on Depth-Image Pose Recognition and Kinematic Mechanisms

ZHUANG Qianwei ,
ZHU Yanfei ,
YU Ning ,
HUANG Dezhong ,
YANG Zheng ,
MA Zhigang ,
YUAN Weihao

Expand

(1. Shanghai Tunnel Engineering Co., Ltd., Shanghai 200032, China; 2. Shanghai Urban Construction Tunnel Equipment Co., Ltd., Shanghai 200137, China; 3. Shanghai Shentong Metro Group Co., Ltd., Shanghai 200122, China)

Online published: 2024-12-12

Fold

摘要

为实现管片自动抓取定位，结合1P5R单臂拼装机，开展深度图像视觉传感器的调研，并对双目、时间飞行法和面结构光3D传感器进行对比分析；将视觉传感器获取的深度图像和管片模型进行匹配，得到管片的位置和姿态；开展视觉传感器与管片拼装机的手眼标定工作，获得它们坐标系间的转换矩阵；利用机器人运动学方法实现单臂拼装机的运动学正逆解，将管片位姿逆解出抓取管片需要的6轴行程，并完成从目前6轴行程到达目标行程的轨迹规划算法。研究和室内试验结果表明： 1）采用深度图像视觉传感器并经过标定可以有效获取管片的位姿； 2）采用运动学正逆解可以利用位姿信息有效引导拼装机，实现管片的自动抓取。

关键词： 深度图像视觉传感器; 管片拼装机; 运动学正逆解; 轨迹规划算法; 转换矩阵

本文引用格式

庄欠伟 , 朱雁飞 , 于宁 , 黄德中 , 杨正 , 马志刚 , 袁玮皓 . 基于深度图像位姿识别和机构运动学的自动抓取管片试验[J]. 隧道建设, 2024 , 44(11) : 2133 -2138 . DOI: 10.3973/j.issn.2096-4498.2024.11.003

Abstract

To realize the automatic capture and positioning of segments, a depth-image vision sensor is explored based on a 1P5R single-arm erector. The performance of a binocular three-dimensional (3D) sensor and a planar-structured light 3D sensor is compared and analyzed. The depth images captured by the vision sensor are matched with a segment model to determine the segment position and attitude. The hand-eye calibration of the vision sensor and segment erector is conducted to obtain the conversion matrix between their respective coordinate systems. The forward and inverse kinematics of the single-arm erector are derived using the robotic kinematics method. Six-axis travel required for grasping segments is realized by inverting the position and attitude of the segment, and a trajectory-planning algorithm is developed from the current six-axis travel to the target travel. Research and laboratory experiments reveal the following: (1) Position and attitude can be effectively obtained using the depth-image sensor and calibration. (2) The forward and inverse kinematic solutions can effectively guide the erector to realize automatic segment grasping using the position and attitude information.

Key words： depth-image vision sensor; segment erector; forward and inverse kinematic solutions; trajectory-planning algorithm; transformation matrix

Options

文章导航

〈

〉

模态框（Modal）标题

摘要

本文引用格式

Abstract