Online feedback hand-eye calibration method based on weighted least square method

Abstract

The invention discloses an online feedback hand-eye calibration method based on a weighted least square method. The calibration method comprises the steps of carrying out the online obtaining of calibration information, determining the position of a calibration plate through automatic analysis, guiding a mechanical arm to move, and extracting the calibration information of hand-eye calibration; weighting the two-step method hand-eye calibration through Parzen window estimation, and solving the two-step method hand-eye calibration by using a least square method; and finally, performing error calculation on the calibration result, accurately judging whether the calibration result meets the calibration precision or not according to the reconstruction distance error, and if not, returning to continuously obtain the calibration information online until the calibration result meeting the precision is obtained. According to the calibration method, the calibration information can be automatically collected, the influence of manual intervention on the calibration result is reduced, the accurate and reliable hand-eye calibration result is obtained through online feedback, and the hand-eye calibration precision is improved.

Description

technical field [0001] The invention relates to the technical field of robot hand-eye calibration, in particular to an online feedback hand-eye calibration method based on a weighted least square method. Background technique [0002] Robots and machine vision are more and more widely used in various fields, especially in the electronics industry, which puts forward higher requirements for the precision and speed of robots. As the basis of vision robot, hand-eye calibration determines the accuracy of the whole vision robot system. Online hand-eye calibration is the online estimation of the relative position and orientation of the end effector or gripper of the robot relative to the camera installed on it. The method of obtaining calibration plate information for the range of motion is not applicable. [0003] The traditional calibration method is to obtain the hand-eye relationship by solving the homogeneous transformation equation AX=XB, where A represents the motion of th...

AI Technical Summary

Problems solved by technology

At present, the commonly used calibration methods are mainly divided into two-step method and simultaneous solution method. The Tsai-Lenz method is a classic two-step solution method. The matrix is ​​decoupled and represented by rotation and translation. The rotation is first calculated, and then the translation is solved according to the result of the rotation. The derivation process is simple and the calculation efficiency is high, but the error of the rotation matrix result is large, and the collection of calibration information will affect the accuracy of the calibration result

At the same time, the accuracy of the solution method is relatively high, but the nonlinear solution is very time-consuming, and it is not suitable for online calibration

Images