Research On Pose Recognition And Intelligent Grasping Technology Of Industrial Parts

Intelligent grasping is an important aspect of robot applications,it was used in pharmaceutical,medical,food processing,manufacturing,household services and other fields widely.In recent years,robotic arms have been rapidly developed in industrial applications.Howerer,the working method is mainly teaching or human-machine collaboration,and the degree of intelligence is not mature enough.In the process of grabbing RJ45 interface type parts by industrial robot,it is necessary to manually place the parts in the specified position according to the specified posture,and grab the parts in a fixed order by the robot arm.Otherwise,the part pins may be caused during the grabbing process.This kind of production method not only wastes labor resources,but also has low production efficiency.There is an urgent need to develop a technology to automatically recognize the position and pose of parts.Guiding the end effector of the robotic arm to adjust the grasping angle to achieve grasping of parts at any position.Based on the above application requirements,this paper designs a binocular vision robotic arm system architecture,which is based on pose recognition and hand-eye control technology to achieve special grasping requirements for scattered parts and improve the intelligent grasping ability of industrial robotic arms.Aiming at the process of industrial automation assembly line robot grabbing RJ45 parts,this article designed to realize the intelligent grabbing of batch parts with binocular camera and robotic arm.To solve the problem of low production efficiency and economic benefits due to manual placement.Hand-eye calibration realizes the unification of the coordinate system of the robotic arm and the binocular camera.Segmenting multiple parts through image processing and recognizing the position and posture of parts with computer vision technology.Designing the grasping path of the robotic arm according to the pose information to avoid collisions with other parts during the process of grasping parts by the robotic arm.Experimental results show that the robotic arm can adjust the grabbing angle to grab RJ45 parts one by one which placed in the workbench with any position.The detection and pose recognition of each part takes less than 150 ms.The grasping accuracy and time efficiency can meet the requirements of industrial robotics grasping applications,effectively solve practical industrial problems.The main contributions of this paper are as follows:(1)Designed a robot arm intelligent grasping scheme based on binocular vision to make the robotic arm perceive the information of the working environment.The binocular camera realizes the detection and positioning of each part.The hand-eye system calibration realizes the universal coordinate of the robot arm and the camera.Transmiting the information observed by the binocular camera to the robotic arm.Experimental results show that without manual involvement,the robot arm can accurately locate the position and pose information of the parts according to the system scheme,and accurately grasp each part according to the plan.(2)A three-dimensional posture recognition method is designed to guide the end effector of the robotic arm to adjust the grasping angle according to the posture of the part to avoid damage to the part during the grasping process.There are two processes for gesture recognition.Firstly,performing 3D models of parts and automatically generating projection images at various angles for network model training in the posture recognition process.Secondly,Yolov3 is used to identify the type of part observation surface,which is used as a priori information to calculate the azimuth angle through the normalized cross-correlation matching method.Using above information to guide the robotic arm to adjust the grasping angle.