An Optimized Depth Complementation of Transparent Objects Based Robotic Arm Grasping System

In this paper, we propose a method to implement a robotic arm for grasping transparent objects and apply it to the grasping of transparent test tubes. Test tubes are one of the frequently used experimental equipment in the chemical industry, and many steps in the experimental process require the use of test tubes to hold reagents. However, as a transparent object, the test tube has unique visual characteristics, which makes it difficult for general-purpose RGB-D cameras to capture its complete depth information. To solve this problem and improve the grasping quality, we propose a robotic arm grasping system using depth completion combined with point clouds. Specifically, we propose a depth learning method to complement the original depth image of transparent objects. In addition, the coordinate transformation relationship between the camera and the robotic arm is obtained by a hand-eye calibration system, while the grasping is performed based on a point cloud map generated from the complementary depth image. Experiments show that our method can significantly improve the depth complementary performance of the transparent object images and achieve accurate grasping by the robotic arm.