A dataset for hand-eye calibration evaluation

Description:
This dataset aims to assess the accuracy of hand-eye calibration methods (i.e., estimation of the transformation between a robot end effector frame and a camera mounted on it). It contains two sets of images and corresponding robot hand poses. The first one (calib_test) contains images of a calibration pattern to estimate the hand-eye transformation. The second one (spirit_reconst) contains images of a pattern to be 3D reconstructed and manually annotated 2D feature points on the images. By performing multi-view 3D reconstruction on the second set and checking the flatness of the reconstructed points, the calibration accuracy can be assessed. The dimension of the calibration pattern in this dataset is 32 mm.

Paper:
Kenji Koide and Emanuele Menegatti, General Hand-Eye Calibration based on Reprojection Error Minimization, IEEE Robotics and Automation Letters/ICRA2019

描述：
本数据集旨在评估手眼标定（hand-eye calibration）方法的精度，即估计机器人末端执行器坐标系与安装于其上的相机之间的变换关系。该数据集包含两组图像及对应的机器人末端手爪位姿。第一组（calib_test）包含用于估计手眼变换的标定板（calibration pattern）图像。第二组（spirit_reconst）包含用于三维重建的标定板图像，且图像上已完成手动标注的二维特征点。通过对第二组图像执行多视图三维重建（multi-view 3D reconstruction）并核查重建点的平面度（flatness），即可评估标定精度。本数据集中标定板的尺寸为32毫米。

论文：
Kenji Koide与Emanuele Menegatti，《基于重投影误差最小化的通用手眼标定》，IEEE机器人与自动化快报/2019年国际机器人与自动化会议（ICRA）

搜集汇总

背景与挑战

背景概述

该数据集旨在评估机器人手眼标定方法的准确性，包含用于标定变换估计的图像与机器人手姿态数据，以及用于3D重建和精度验证的带标注点的图像。数据集基于32 mm标定板，支持通过多视图3D重建检查点平面度来评估标定精度，相关研究发表于IEEE Robotics and Automation Letters/ICRA2019。