Mechanical arm grabbing system and method based on edge feature matching

Abstract

The invention discloses a mechanical arm grabbing system and method based on edge feature matching, and the system comprises a central computer, a mechanical arm grabbing module, and a point cloud image collection module, and the central computer comprises a format conversion module, a feature extraction module, a pose estimation module, and a hand-eye calibration module. The method comprises twostages of offline modeling and online matching, a central computer carries out offline modeling to form model point cloud and carries out feature extraction, a point cloud image acquisition module forms corresponding scene point cloud information by acquiring a workpiece image and carries out an online matching process with the model point cloud, so that the pose of a workpiece is determined; andfinally, the mechanical arm grabbing module carries out the grabbing process. The invention can be applied to subsequent workpiece sorting only by performing offline feature modeling once, and uses edge points and fitting tangent points of the edge points to features, so that the influence of the normal is reduced, the sorting effect of the workpieces is improved, the pose estimation time is shortened by using a two-dimensional image, and pose estimation and grabbing of stacked workpieces are realized.

Description

technical field [0001] The invention relates to a grabbing system and method of a robotic arm based on edge feature matching, and belongs to the technical field of intelligent grabbing of robotic arms. Background technique [0002] With the gradual deepening of the transformation of the manufacturing industry to intelligence, 2D vision has been difficult to meet the application requirements of space capture, and the application of 3D vision technology in the industrial production process has gradually matured. However, the problem of autonomous grasping of the manipulator has always been the main bottleneck restricting the development of industrial robots and service robots, and the pose recognition and positioning of the grasping target is a necessary prerequisite for the realization of the autonomous grasping of the manipulator. At present, most robots in the industry are still in the stage of trial teaching, unable to make effective judgments on complex environments, espe...

AI Technical Summary

Problems solved by technology

At present, most robots in the industry are still in the stage of trial teaching, unable to make effective judgments on complex environments, especially in the case of densely stacked workpieces, it is difficult to achieve effective object recognition, and using classic point-to-point features as global point cloud features In the case of the descriptor, when it is applied to some workpieces that can extract normal features, it will cause inaccurate matching. In addition, if the scanned point cloud has a large specification, and due to the time complexity of the algorithm The relationship will cause the computer calculation time to be too long, which will seriously affect the grabbing efficiency of the industrial robotic arm

At present, there is no system and method that can realize the accurate identification and grasping of workpieces by industrial manipulators in complex environments, while ensuring the grasping efficiency

Images