Research On Grasping Method Of Robotic Arm Based On 3D Vision Positioning

With the development of technology and the transformation of traditional manufacturing,robot intelligent grasping has been widely used in industrial production.Combining industrial cameras with the robot is a significant trend in the development of robot in the future.At present,the common 2D vision relies on the shape and color of the object for recognition,and it can only obtain the plane image of the object.So,it cannot meet the complex operating environment.3D vision can not only directly obtain the depth information of objects,but also recognize objects without texture and occlusion.In this paper,we build a robot grasping system based on 3D vision based on the robot control system ROS.Firstly,the construction of the vision-based robot grasping system is completed.We carried out hardware selection,and then used Zhang’s calibration method to obtain the parameter model of the camera.Then we use the Tasi two step method to perform hand-eye calibration of the camera and robot.The system of this paper chooses the matching method of eyes and hands.Secondly,the specific steps of target recognition and localization are introduced in detail.We model the target object in 3D by stitching the residual defect cloud into a complete template point cloud.We preprocess the scene point cloud using methods such as down sampling,plane segmentation and filtering.We extract point-to-point features from the processed scene point cloud and template point cloud,and then perform point cloud registration.In the registration process,the method of Hough sampling and pose clustering is used in this paper to improve the registration accuracy.Finally,the optimal path planning is carried out for the robot to complete the grasping work.In this paper,the robot operating system ROS is used to control the grasping system.By establishing a robot model,real-time monitoring of the grasping scene is realized.We use the fast-exploration random tree method to generate the grab path quickly,and then select the path with the shortest travel as the optimal path.At the same time,we plan the grasping angle of the grasper.