图像信息输入下的软线重构与抓取验证  

Flexible wire reconstructing and gripping verification under the image information input

作　　者：邱介禄 王明明[1,2] QIU Jielu;WANG Mingming(School of Aeronautics,Northwestern Polytechnical University,Xi’an 710072,China;Research&Shenzhen Research Institute of Northwestern Polytechnical University,Shenzhen 518057,China)

机构地区：[1]西北工业大学航天学院,陕西西安710072 [2]西北工业大学深圳研究院,广东深圳518057

出　　处：《智能系统学报》2025年第2期435-444,共10页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(61973256);广东省基础与应用基础研究基金项目(2024A1515011178).

摘　　要：针对使用机器人对柔性物体进行操作的复杂任务,设计了一种简洁高效的软线抓取方法。首先利用双目相机和图像处理方法对软线进行测量,结合坐标系间的转换关系,从图像中还原了其特征点的空间坐标信息;采用Frenet标架对软线进行了建模重构,以获取软线上任意点的位置与弯曲方向信息;提出了六自由度机械臂的位姿表示及转换方法以及相机的标定流程,将曲线任意弧长处的坐标值和Frenet标架方向转换为夹爪末端的位姿参数,最后开展了目标点处沿软线弯曲方向抓取软线的地面实验,验证了方法的有效性。To address the complex task of operating flexible objects with a robot,a concise and efficient method for grasping soft wires is designed.First,the soft wire is measured using a stereo camera and image processing techniques.By combining the transformation relationships between coordinate systems,three-dimensional coordinate information is reconstructed from the image data.The soft wire is then modeled and reconstructed using the Frenet frame,enabling the acquisition of positional and bending direction information at any arbitrary point along the wire.The representations and transformation methods for the six-DOF robotic arm pose,along with the camera calibration process,are provided.These enable the conversion of the Frenet frame’s coordinates and directional data for any point along the curve into the robotic gripper’s pose.Finally,a ground experiment was conducted to demonstrate the robot’s ability to grasp the soft wire along the bending direction of a target point,verifying the feasibility of this method.

关 键 词：软线 视觉感知 深度相机 图像处理 坐标计算 曲线重构 位姿转换 机械臂抓取 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]