Research on Autonomous Grasping Technology Based on Vision

The grasping operation is limited by visual observational conditions, operator level, etc., in the process of robot teleoperation. In response to this problem, this paper proposes a method of autonomous grasping based on vision. A depth camera is carried by the robot to obtain the color and depth information of the environment. The operator selects the target with a box on the monitor to determine the approximate range of the target. Combining the target box, depth information and color information, the image foreground is extracted to obtain the plane contour and depth information of the target. Then, the pose data of the target are calculated, and the grasping position is estimated and converted into a position in the robot coordinate system. After the grasping position is determined, the grasping trajectory of the manipulator is planned and optimized, combined with the environmental obstacle information. Finally, the manipulator is controlled to complete the grasping process. Experimental results show that the autonomous trajectory planning of the robot is continuous and controllable and that the grasping process can be completed efficiently with high accuracy in the recognition and positioning processes.