基于三维视觉的随机工件识别与姿态估计

3D Vision Based Object Detection and Pose Estimation

针对堆叠物品的识别与定位,本文提出了一种自适应动态阈值分割与分层识别方法,实现了对随机、无纹理堆叠物品的识别。实验结果表明用该方法可以准确识别随机堆放工件中的目标,目标工件在X,Y,Z轴方向上的平均定位误差为1.01mm、0.99mm、1.91mm。

Aiming at the detection and pose estimation of stacked items,this paper proposes an adaptive dynamic threshold segmentation and layered recognition method to realize the recognition and localization of random and untextured stacked items.The experimental results show that the method can accurately identify the target in the random stacking workpiece,and the average positioning error of the target workpiece in the X,Y and Z directions is 1.01 mm,0.99 mm,1.91 mm.