Research On Target Pose Estimation And Grasping Method Based On Line Feature

Machine vision technology provides rich control information for robot work,making it more flexible and adaptable to production needs in complex environments.As one of the key technologies of robot grasping with vision,pose estimation has attracted the attention of many scholars.According to whether the surface of the target is affixed with a logo,it is divided into cooperation goals and non-cooperation goals.The research object of this paper is the non-cooperation goals known by the model.Aiming at the artificial plane target and the three-dimensional target with weak texture information,this dissertation makes a deep research on the pose estimation method based on the linear feature,and proposes a planar target pose estimation algorithm based on line correspondence and weighted least squares,Filtering and weighted least square three-dimensional target pose estimation algorithm,and set up an IRB1200 6-DOF robot based on the workpiece positioning and grasping system.Specific research includes:Firstly,the edge line is selected as the feature of pose estimation due to the rich feature line and the more reliable line detection for pose estimation of weakly-textured targets.By comparing the performance of linear extraction algorithm based on Hough transform and LSD linear detection algorithm,we choose LSD algorithm to extract image line.Secondly,a pose estimation algorithm based on linear correspondence and weighted least squares is proposed for pose estimation problem of planar target,.In order to overcome effectively the influence of end-point positioning error in image segment,this paper presents a new pose-error function based on three-dimensional space,which integrates the information of line endpoint and midpoint.Based on the pose error function presented in this paper,the linear distance from the line to the model projection is minimized.The partial derivative matrix of the error function is deduced.The problem is transformed into the weighted least squares problem,and iteration is added to ensure the accuracy of the solution.In the planar workpiece positioning experiment,the pose estimation algorithm based on the straight line still shows higher accuracy when the initial pose is inaccurate.Thirdly,an initial pose estimation algorithm based on particle filter is proposed for the problem of pose estimation and exact grasping of three-dimensional target in complex environment.Based on the initial pose obtained by particle filter,weighted least square method is used to solve the target position Pose parameters to improve the efficiency and accuracy of the algorithm.Due to the large number of'three-dimensional target lines,it is easy to cause a mismatch between the model line and the image straight line.Therefore,this paper introduces robust estimation to robustly estimate pose parameters,and eliminates the influence of the line mismatch between model line and image.Finally,the pose and position estimation algorithm based on particle filter and least squares is successfully applied to the pose tracking of moving targets.The algorithm is applied to the experiment of robotic grasping workpieces.Experiments show that the algorithm has high precision and good robustness.In this paper,the pose estimation method based on linear features is studied in depth.The proposed method can also be used for other tasks such as aircraft aiding,augmented reality and scene matching.