摘要
为了获取生产线上无序且被遮挡的工件的准确位姿,为后续机械臂对工件特定部位的抓取提供支持,本文使用高精度三维激光扫描仪扫描生产线上无序摆放的工件,然后对原始点云先后采用Voxel-Grid降采样、平面分割和DBSCAN聚类,分割出单个工件点云,再用K-4PCS粗配准和kd-tree加速的GICP精配准方法,将工件实际点云与标准的点云模板(由工件的机械设计图转化而来)进行配准,最终得到了两者之间精确的位姿变换矩阵,进而获得工件摆放的准确位姿。为了验证论文提出方法的有效性,对生产线上3组形状不同工件的摆放位姿进行了测试。结果表明,该方法对于不同形状的工件摆放位姿都有很好的求解精度(误差≤3 mm)。
To obtain the accurate position of the disordered and obscured workpieces on the production line,and to support the subsequent gripping of specific parts of the workpieces by the robot arm,a high-precision 3 D laser scanner is used to scan the disordered workpieces on the production line in this paper.The Voxel down-sampling,plane segmentation and DBSCAN clustering methods are applied to the original point cloud to segment the individual workpiece point clouds.Then,the coarse registration algorithm of k-4 PCS and the precise registration method of the GICP based on accelerated kd-trees are adopted to register the actual point cloud of the workpiece with the standard point cloud template(converted from the mechanical design drawing of the workpiece).Finally,the accurate posture transformation matrix between the two is obtained,and the accurate posture of the workpiece placement is obtained.In order to verify the effectiveness of the proposed method,the position and attitude determination experiment is carried out on three workpieces in different shapes in the production line.The results show that the proposed method has good solution accuracy(error≤3 mm)for different shapes of workpiece placement.
作者
黄磊
朱秀敏
石雨婷
曹恒宣
HUANG Lei;ZHU Xiu-min;SHI Yu-ting;CAO Heng-xuan(College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 210037,China)
出处
《激光与红外》
CAS
CSCD
北大核心
2023年第7期1026-1032,共7页
Laser & Infrared
基金
国家自然科学基金资助项目(No.31901239)
2021年江苏省大学生创新训练计划(No.202110298016Z)资助项目。