Research On Position And Pose Recognition Technology Of Randomly Stacked Bars Based On Point Clouds

On the production line,the traditional feeding mechanism occupies a large area,has a lot of noise,and requires the participation of workers.With the development of industrial intelligence,the traditional feeding method no longer meets the demand for automation.At this stage,most of the smart chemical factories adopt a two-dimensional camera guidance method,which is unable to perform good recognition of stacked objects that obstruct each other,and the accuracy and generalization ability are low.In order to solve the above problems,this paper constructs a visual recognition system based on point clouds for randomly stacked bars with different aspect ratios,focuses on the research of pose recognition technology,builds a hardware platform,and designs a robot grasping experiment for verification.The intelligence of the feeding system.This article first analyzes the stacking characteristics of bars with different length to diameter ratios,and proposes an overall system plan based on actual needs.The main structure of the system scheme is analyzed from the visual recognition module and the robot grasping module,and the system flow is designed.Under the ROS system,the calibration of the vision system is completed based on the Zhang Youzheng calibration method.Secondly,by analyzing the requirements of point cloud preprocessing,including removing irrelevant point clouds,noise and improving the quality of point clouds,this article compares and optimizes the point cloud preprocessing methods for this scene;by analyzing different point cloud segmentation principles and effects,Improved the original segmentation algorithm to accurately and efficiently segment the point cloud of the bar to be grasped;by extracting different features of the point cloud,the segmentation point cloud was identified and classified;the random consistent sampling method was used,and the slicing method was combined with The mean-shift method recognizes the pose of the bar;finally,the feasibility of the algorithm is verified by a simulated grabbing experiment.In response to the above research,an experimental platform was built to verify whether the pose recognition system meets the requirements.Choose EPSON 6-axis robot;design flange connectors and end grippers according to application scenarios;establish communication between vision system,robot system and computer through TCP/IP protocol;convert the recognized pose information into robot format by solving transformation matrix Pose,and determine the layout of the experimental equipment through the integrated robot software simulation;through the form of the target ball,based on the ICP calibration of the hand,the conversion matrix from the camera coordinate system to the robot coordinate system is solved;Grabbing experiments were carried out on the scene of materials,which verified the feasibility and reliability of the system.