Research On Key Technologies Of Object Detection And Grasping Based On Visio

At present,the grasping technology of the desktop robotic arm has become a hot issue in the field of robot research.The desktop robotic arm can realize human-machine cooperation and help people complete some repetitive,high-risk or high-precision tasks to reduce labor intensity,reduce work risk,improve work efficiency and accuracy.Therefore,the research on the grasping of household items by the desktop service robotic arm has important practical application value.This topic researched the key technologies encountered in the grasping of desktop robotic arm in family life,and is committed to building a set of low-cost and high-efficiency robotic arm grasping systems.Firstly,this article analyzed the system requirements and designd the system architecture and software platform to build a reliable robotic arm visual grasping system.This system could use external environmental information to detect objects and plan obstacle avoidance paths to achieve autonomous grasping.Analyzed and studied the kinematics theory of the robotic arm,used the D-H method to establish the mathematical model of the UR5 robotic arm,performed forward and reverse kinematics analysis,and in the ROS simulation environment,used Move It! to configure the robotic arm to provide for the grasping simulation of the robotic arm necessary foundation and platform.In order to reduce the amount of calculations and parameters of the target detection algorithm and ensure the detection accuracy of household objects,this paper improved on the basis of the YOLOv5 m algorithm,and used the lightweight module Ghost Bottleneck to replace the Bottleneck in the network backbone C3＿X structure.The CBAM module was embedded in the YOLOv5m＿G network before the detection head.It had been verified by experiments that,under the premise of ensuring the detection accuracy,the parameter amount of the original YOLOv5 m algorithm was reduced by 30%,and the calculation amount was reduced by 37%,which effectively reduced the parameter amount and calculation amount.In addition,using a CNN based grasping detection algorithm,a fiveparameter grasping mathematical model was selected to determine the object grasping frame,so as to realize the precise grasping of the specified target object by the robotic arm.The collision detection and path planning algorithms of the robotic arm were studied,and the performance of the RRT series path planning algorithms was compared through simulation analysis.The Informed-RRT* algorithm was used to realize the trajectory planning of the robotic arm,and the cubic Bezier was used to smooth the movement path of the robotic arm to reduce wear and damage and obtain a more optimized path.Finally,the platform of the simulation experiment was built.Two voice control modules for human-computer interaction and robotic arm gripping were designed.The robotic arm grasping experiment was carried out in the ROS simulation environment,which verified the effectiveness of the functions of each part of the service robotic arm grasping system.The final experimental results proved that it can meet the established requirements.