Hand-eye Calibration And Object Localization For Industrial Robotic Application

In order to meet the requirements of the transformation and upgrading of manu-facturing industry, the intelligent industrial robot with vision system is used more and more widely in modern factory. In order to adapt to the rapid deployment of production system in flexible manufacturing and meet the operation requirement of accurate object recognition and localization, this thesis researches and explores the issue about indus-trial robot hand-eye calibration and object recognition and localization. The research contents and results of this thesis mainly include the following aspects:1. An online autonomous hand-eye calibration system has been designed and imple-mented. While performing calibration algorithm, the system can automatically collect the calibration data. The calibration object motion space planning based on the camera imaging model ensures that the calibration object appears in the camera field of view when collecting data, so that the system can obtain effec-tive calibration data. Hand-eye calibration calculation adopts linear algorithm to ensure the real-time online computing. At the same time, the effective system flow is designed to control the beginning and end of the calibration and to en-sure adequate calibration data to eliminate the influence of the observation error. Experiments show that the autonomous hand-eye calibration method designed in this thesis can obtain convergent calibration results, and the whole calibration process takes only 15min.2. An optimized hand-eye calibration algorithm based on minimizing the reprojec-tion error has been proposed and implemented. In this algorithm, the camera imaging model and robot hand-eye model is regarded as a whole model, and the image coordinates of corner points on the checkerboard are adopted as di-rect observation data. In the pixel space, model parameters are optimized and the estimation error of the hand-eye transformation matrix is converted to the reprojection error of corner points on the checkerboard. The optimization goal is minimizing the reprojection error. In order to solve the optimization problem with two unknown parts, the iterative optimization method is adopted. Experi-ments show that the proposed algorithm can achieve the calibration accuracy of 0.873mm.3. An object recognition algorithm based on ORB(Oriented FAST and Rotated BRIEF) and a localization algorithm based on local shape feature have been designed and implemented. In this algorithm, the object recognition algorithm based on fea-ture points matching is adopted. Because of excellent rotational invariance and real-time, ORB feature has been adopted. The homography matrix between the perceptual image and the template image calculated by RANSAC is used to com-plete initial localization. On the basis of this, an optimized localization algorithm based on local shape feature is proposed to relocate the object. Experiments show that the recognition algorithm can achieve fast and stable recognition of the cir-cuit board object in the industrial environment, and the optimized localization algorithm can reduce the relative localization error from 0.5658mm to 0.1770m-m.