基于VGG16-RGBD的散乱件抓取场景下分拣点检测

Grasp Point Detection of Bin Picking Based on VGG16-RGBD Network

针对在密集环境下的散乱件抓取(Bin Picking)问题,由于存在大量的遮挡,所以要求机器人能够在有遮挡情况、物体杂乱放置的环境中对未定义物体进行可靠的抓取点检测。本文提岀了一种新颖的基于VGG16-RGBD网络的抓取点检测方法,在密集环境下提高了机器人抓取的准确率和精度。通过在真实机器人上实现抓取动作,证明了此方法的有效性,表明了此方法可以准确检测物体的抓取点,并且在复杂环境中达到了 94%的成功率。

This paper focuses on the Bin Picking problem in dense environment. Because there are a lot of occlusions, the robot is required to detect undefined objects reliably in the environment where there are occlusions and objects are placed in disorder. In this paper, a novel grasping point detection method based on VGG16 - RGBD network is proposed, which improves the accuracy of robot grasping in dense environment. This paper proves the validity of this method by realizing grasping action on a real robot. It shows that this method can accurately detect the grasping points of unknown objects, and achieves 94% success rate in complex environment.