A Robot 3D Grasping Application Based on Binocular Vision System

3D vision system plays an essential role in robot grasping applications in complex environments. Six-degree-of-freedom pose estimation for weakly textured targets is a current research hotspot in the field of machine vision. This paper proposes a method based on feature region distribution to optimize the feature matching point pairs. This paper presents a global feature constraint-based target pose estimation method, enabling the visual system to filter the target to grasp complex backgrounds and stacking situations. Finally, we validate the effectiveness of the proposed approach through robot grasping experiments.