Robot automatic grabbing system and method based on 3D vision

Abstract

The method comprises the steps that a grabbing robot sends a request data instruction to an industrial personal computer, the industrial personal computer receives the request data instruction and judges whether a processing result is cached or not, if yes, the processing result is returned to the grabbing robot, and if not, a 3D camera is controlled to take a picture; the industrial personal computer receives a photographing instruction of the grabbing robot and then controls the 3D camera to photograph; the industrial personal computer receives the original three-dimensional point cloud data and judges whether a target part is shot or not; if the target part is shot, point cloud processing calculation is carried out based on the original three-dimensional point cloud data to obtain part space pose data and state information, and the part space pose data and the state information are cached as processing results; if the target part is not shot, judging whether all the parts are shot or not to obtain part space pose data and state information, and caching the part space pose data and the state information as a processing result; and the grabbing robot grabs the parts according to the processing result and outputs a photographing instruction and a material frame replacing instruction. The manipulator has the advantages of being high in grabbing success rate and fast in takt.

Description

technical field [0001] The invention relates to the technical fields of industrial robots and machine vision, in particular to a 3D vision-based robot automatic grasping system and method. Background technique [0002] The traditional robot application is to control the robot to complete the predetermined command action through the teaching pendant. However, when the working environment and tasks of the robot change, the robot often cannot adapt to these changes quickly, which greatly limits the application scenarios and application efficiency of industrial robots. [0003] With the development of machine vision technology, machine vision technology is gradually applied to industrial scenarios. However, traditional two-dimensional vision can only be applied to the recognition and positioning of planar objects, and cannot cope with three-dimensional scenes such as tilting and flipping of parts. [0004] 3D cameras can obtain object space point cloud coordinates to make up fo...

AI Technical Summary

Problems solved by technology

However, traditional two-dimensional vision can only be applied to the recognition and positioning of planar objects, and cannot cope with three-dimensional scenes such as tilting and flipping of parts

[0004] 3D cameras can obtain object space point cloud coordinates to make up for the shortcomings of traditional 2D vision, but the large amount of 3D point cloud data, complex processing, and complex communication logic with robots greatly limit the application of 3D vision

However, this method uses binocular depth cameras and RGB cameras instead of directly using 3D point cloud data, and the method based on deep learning is not robust; the Chinese invention patent with application number CN202011256951.7 "a robot based on 3D vision Fully automatic disorderly feeding system" based on 3D vision robot automatic disorderly feeding system, including feeding device, unloading device, camera device, movable manipulator and central control system

However, this method needs to pre-collect the type of material and its spatial six-sided characteristics, which is cumbersome and inefficient.

Images