Weak-texture workpiece, and method and system for recognizing and detecting three-dimensional pose

Abstract

The invention discloses a weak-texture workpiece, and a method and a system for recognizing and detecting the three-dimensional pose. After a color image of a target workpiece acquired by an RGB-D sensor and a gradient pattern template library are subjected to image pyramid search based on a similarity evaluation function, a three-dimensional point cloud is built on a depth image according to a matching result, and finally, through filtering and iterative closest point computation on the three-dimensional point cloud, the precise three-dimensional pose of the target workpiece is obtained. An industrial robot can accurately and intelligently recognize the weak-texture workpiece and acquire the three-dimensional pose information, and tasks such as grinding and assembling on the workpiece can be completed.

Description

Technical field [0001] The invention relates to a technology in the field of graphics processing, in particular to a method and system for identifying and detecting weak-textured workpieces and their three-dimensional poses. Background technique [0002] With the continuous development of machine vision technology, industrial robots can rely on intelligent vision systems to automatically and accurately complete tasks such as workpiece grasping, processing, polishing, and assembly. In the past, these tasks often required workers to teach and assist the robot. To be completed well, the introduction of machine vision technology will greatly improve product quality and production efficiency. The purpose of workpiece recognition and 3D pose estimation is to recognize workpieces of different types and structures and obtain their 3D position and pose information relative to the visual sensor, which is used to guide industrial robots to automatically complete the grasping, processing, gr...

AI Technical Summary

Problems solved by technology

[0003] The existing weak-textured workpiece recognition and 3D pose estimation methods have the following technical problems: 1) The workpiece recognition based on 2D images is highly suitable for application scenarios, and it can only be used in factories to capture and analyze workpieces on the assembly line. At the same time, it is also required that the background of the conveyor belt and the workpiece have a certain contrast, otherwise it is difficult to meet the requirements

The three-dimensional pose information of the workpiece required for robot grinding or assembling parts cannot be accurately obtained; 2) Realizing workpiece recognition based on a 3D camera, data acquisition and processing are more complicated, and the amount of calculation is large, and most of them are only applicable to bin-picking scenarios 3D pose acquisition of a single workpiece

Images