Research On Visual Servo Tracking Docking Technology Of Multi Degree Of Freedom Module Transfer Mobile Platform

The robot serves as the "intermediate gear" link in the "Industry 4.0" system.It involves the development of many control technologies.The end effector based on visual feedback automatically docks the workpiece to the designated station accurately,which is a research topic of great significance.Especially in the fields of aerospace rendezvous and docking,industrial clamping and automatic grabbing of target objects,visual servoing technology is widely used.It can not only guarantee the servo characteristics of the mechanism,but also fulfill the requirements of the docking task excellently.This article focuses on the elaboration and research of the mechanism kinematics,image processing,and servo motion system principles involved in the dynamic tracking and docking process of the multi-degree-of-freedom mobile docking platform.The main content of the article includes the following aspects.In view of the fact that the multi-degree-of-freedom module transport mobile platform is composed of a Mecanum wheel chassis platform and a 3-SPS/PU parallel level adjustment mechanism in series,this paper studies the properties of kinematics and differential kinematics.The parallel kinematics equation between the 3-SPS/PU level adjustment platform and the state of its driving mechanism is constructed by using the vector closed-loop method.At the same time,the differential kinematics equation of the mecanum wheel chassis plane adjustment platform is established.Based on this,the relationship between the six-degree-of-freedom pose change of the end-effector and the state of each driver of the hybrid mechanism is further discussed.Finally,simulation is used to verify the correctness of the kinematics equations.In different initial poses and motion amounts,the changes in the state of the drive conform to the motion conditions of the mechanism under real working conditions.The image target tracking algorithm combined with improved Kalman filter and Mean Shift is studied.First,extract the color histogram in the RGB space of the target image as the matching template for tracking.The Mean Shift algorithm is integrated into the Kalman filter target tracking posterior estimation result optimization.Improve the robustness of the target tracking algorithm.Among them,for the problem of fewer target tracking features,the image gradient histogram is fused into image matching.At the same time,the size and update of the target tracking area are periodically changed according to the change of the pixel distance between the feature points.Finally,it is verified on the OTB-100 data set,and the tracking result curve of MSFT and KF shows that the improved algorithm has higher tracking accuracy and robustness.The long base distance binocular space pose estimation algorithm is discussed.First,the image is filtered,image enhanced and other pre-processing to ensure the extraction of feature points.Then,the sub-pixel corner extraction algorithm of improved centroid method is used to optimize the normal corner points that have been obtained.Experiments prove that the improved algorithm is better than the traditional sub-pixel corner extraction algorithm in terms of speed and floating point calculation performance.Next,the SVD decomposition is used to extract the six-degree-of-freedom spatial pose of the monocular camera,and after Rodrigues transformation,the final pose of the plane where the longbase-distance binocular camera is located is obtained.Finally,in order to effectively use the image pose feedback information obtained above,according to the actual docking situation of the multi-degree-of-freedom module transfer mobile platform,the actual docking pose is moved forward in the positive direction of the Z axis.At the same time,it focuses on the principle and construction of the PBVS(position based visual servoing)visual servoing system of the multi-degree-of-freedom module transfer mobile docking platform.In order to verify the effectiveness of the above improved algorithm.A long base distance binocular vision pose estimation platform was built,and the improved Kalman filter target tracking algorithm(IKMFT),the pose estimation algorithm and the sub-pixel corner extraction algorithm of the improved centroid method were verified.At the same time,the characteristics of the PBVS visual servo system are simulated,and the posture change curve of the end effector in space is obtained.Finally,the simulation results show that the improved and optimized algorithms proposed in the previous chapters have good performance in timeliness and accuracy,and can be practically applied to engineering.