堆叠筒子纱机器人无序分拣方法

Robotic Unordered Sorting Method for Stacked Bobbin Yarns

针对筒子纱分拣机器人对作业目标堆叠时识别不准确、抓取位置不稳定的问题,课题组提出了一种堆叠筒子纱的机器人无序分拣方法。构建了具有3D视觉感知的筒子纱分拣机器人系统,利用Kinect V2相机获取堆叠筒子纱的图像信息,对采集到的筒子纱点云进行处理;通过M估计抽样一致性算法(MSAC)和直通滤波算法对原始点云进行裁剪,采用改进E-R分割算法和ICP算法完成点云的分割和配准,得到筒子纱的位姿信息;最后使用机器人进行抓取实验。实验结果表明:该方法能够实现对堆叠场景下的作业目标的识别定位,系统的抓取成功率达到86%,可以满足筒子纱分拣的实际生产需求。

In response to the problems of inaccurate recognition and unstable grasping position of the stacked operation targets by the bobbin yarn sorting robot, a disorderly sorting method by robot for stacked bobbin yarns was proposed. The robot system with 3 D vision perception was built, using the Kinect V2 camera to acquire the image information of the stacked yarn, and the point cloud of collected yarn was processed. The original point cloud was cropped by MSAC(M-Estimation Sampling Consistency Algorithm) and straight pass filtering algorithm, and the segmentation and alignment of the point cloud was completed by improved E-R segmentation algorithm and ICP algorithm to obtain the positional information of the bobbin yarn, and finally the robot was used for verification by grasping. The experimental result shows that this method can recognize and locate the working target in stacking scenario, and the success rate of the system reaches 86%, which meets the actual production requirements of bobbin yarn sorting.