Design of an Unordered Gripping System Based on 3D Vision

Shunqiang Chen,Junbo Bai,Liao Xu,Luo Zhen,Feitong Wang,Li Bo,Junjie Bai
DOI: https://doi.org/10.1088/1742-6596/2577/1/012003
2023-09-03
Journal of Physics: Conference Series
Abstract:For the current mainstream 2D vision-guided robotic arm gripping system in industrial production, the vision part is seriously affected by the lighting environment, camera imaging distance, and other factors. It cannot provide more comprehensive 6DOF position information on the target workpiece and complete the gripping of mutually scattered and stacked workpieces. This paper designs a disordered grasping system based on the 3D vision to solve the above problems. In addition, the traditional Euclidean clustering algorithm and fast point feature histogram algorithm are improved to enhance the segmentation and feature recognition efficiency of the point cloud. The accuracy of the system is 94.3 % without stacking and 88.2 % with stacking, which is 5.2 % higher than the average accuracy of the system before the algorithm improvement and more than 2s higher than the time. The accuracy and speed of the system meet the requirements of general industrial production for a robotic arm grasping system, which has certain practical application value.
What problem does this paper attempt to address?