Six-dimensional position-attitude active measurement method of space circle based on monocular hand-eye system

Abstract

The invention discloses a spatial circle six-dimensional position-posture active measurement method based on a monocular hand-eye system. The method comprises the following steps of controlling the monocular hand-eye system to acquire images on at least two different positions, calculating a center coordinate, a normal vector and a radius of a spatial circle according to the acquired images, and obtaining six-dimensional position-posture estimation of the spatial circle, wherein the acquired images on the different positions meet the requirement that a projected oval is located in the central position of each image, and an included angle of optic center ligature of the monocular hand-eye system and a circle center ligature of the spatial circle at different positions is larger than a preset value, so that the posture of a circular workpiece is more accurately calculated.

Description

technical field [0001] The invention relates to a six-dimensional position-posture active measurement method of a space circle based on a monocular hand-eye system. Background technique [0002] Object pose measurement has very important applications in robotics, photogrammetry, and machine vision, especially in the field of robotics. Accurately and quickly obtaining the six-dimensional pose of an object is crucial for robots to grasp objects. This is especially true at the assembly site. Accurate measurement of the position and attitude of the assembly is the key to the rapid grasping of objects by industrial robots, alignment and installation, and can reduce mechanical wear and improve the assembly efficiency of robots. [0003] At present, there are relatively mature methods for non-contact pose estimation of objects using machine vision technology. Some methods exploit the correspondence between 2D image features and 3D models to estimate the six degrees of freedom posi...

AI Technical Summary

Problems solved by technology

Some literatures use RGB-D depth cameras to build object models from images acquired from several different perspectives. Although this method does not require an accurate 3D CAD model, modeling objects from 3D point clouds requires a lot of calculations.

[0004] The stereo vision system does not need to know or establish the 3D model of the object in advance, but the stereo vision system is large in size in order to ensure the distance between the two cameras, so it is not suitable for installation at the end of the robotic arm

[0005] In summary, the existing methods do not consider how to actively control the movement of the monocular hand-eye system by controlling the initiative of the hand-eye system, that is, to improve the accuracy of object pose measurement, in some specific favorable positions. Obtain images of circular workpieces, and calculate the pose of circular workpieces more accurately based on the images collected at these favorable positions

Images