Research On Robot Target Position Gesture Estimation And Grab

With the demand for labor in the service industry grows,robots are increasingly being used in industries and services.Industrial robots on automated production lines,“grabbing-placement” have become the most frequently implemented operations,and in the daily life of service robots,“grab-and-place” is also essential.In order to successfully complete the operation,it is necessary to acquire the target object,that is,the position and posture information of the operated object.Because this paper carries out the robot target position estimation and grab research.The images are acquired by two different visual sensors,a depth camera and a binocular camera,respectively.For the objects with linear features in the scene,the position and posture of the captured objects are estimated by different algorithms.The specific research contents are as follows:First,on the basis of studying the camera imaging model and the principle of visual measurement,the Kinect camera and the binocular camera were respectively calibrated.Secondly,the pose estimation method based on point cloud feature is studied.For the scene image acquired by the Kinect camera,the pass-through filtering and the plane-fitting-based algorithm are used to remove the background cloud of the scene point cloud and the point cloud of the plane of the object to be captured,and the point cloud of the object to be captured is realized by the European clustering algorithm.Separate separately.Then,the ISS3 D feature points of the point cloud of the object to be captured are extracted,and the FPFH feature descriptor is generated,and the feature description submodel of the object to be captured is established.The SAC-IA algorithm is used to match the feature descriptors of the point cloud to be captured with the feature descriptors in the model to obtain a rough pose estimation.Then the ICP algorithm is used to optimize the calculation of the optimal transformation matrix,and the pose information of the object is realized.Accurate estimation.Thirdly,the binocular pose estimation method based on the edge feature of the target is studied.Aiming at the scene image acquired by the binocular camera,a binocular pose estimation method based on the target edge feature is proposed.Firstly,aiming at the object with linear features in the scene,a Hough transform combined with line segment clustering algorithm is proposed to accurately extract the straight edge of the object.Then,the matching and screening mechanism is developed by using the binocular stereo calibration parameters.Matching pairs of straight lines;and designing a method for intersecting lines in space plane,three-dimensional reconstruction of matched line pairs;finally,the initial pose matrix is calculated by EPnP algorithm,and the 3D reconstruction result of the straight edge of the object is combined with ICP algorithm.The registration edge point cloud and the target edge point cloud are finely registered to obtain an accurate rotation translation matrix,thereby realizing accurate position and attitude estimation for objects with distinct linear features.Finally,a robotic gripping experiment with hand-eye coordination was performed.The algorithm is applied to the hand-eye calibration of the Kinova manipulator and the re-projection error analysis is performed.The design comprehensive experiment compares the performance of the pose estimation method based on the point cloud feature and the binocular pose estimation method based on the target edge feature.Finally,the calculation result of the position and posture is taken as the input of the robot arm system,and the grasping of the target object is realized.