Under CopyrightGarcia, C.C.Garcia2022-03-0707.08.20021998https://publica.fraunhofer.de/handle/publica/28984510.24406/publica-fhg-289845The current prototype of the JANUS robot consists of two arms and a neck with two mounted color cameras. Calibrating JANUS means to find the geometric relationships between each of these components and the reference coordinate system of the robot. The calibration procedure that we present in this report is completely vision-based: the relationships between each camera and the neck and between each arm and the neck are determined using visual measurements, which leads to a low-cost and automatic setup. For this purpose, we developed two main algorithms. The first one implements a non-linear optimization method using a quaternions formulation for camera calibration from 2D to 3D point or line correspondences. The second one is a real-time camera pose estimation method based on the iterative use of a paraperspective camera model. Using the results of calibration we show how we solve some classical computational stereo vision and 3D reconstruction problems. Also are provided experimental results for each of the different steps of the complete calibration of JANUS.Contents S.5-6 1 Introduction S.7 2 The hand-eye robot JANUS S.8-9 - 2.1 The Arms S.8-9 - 2.2 The Stereo Vision Head S.9 - 2.3 The brain S.9 3 Notations, relationships and reference frames S.10-13 4 Eye calibration S.14-18 - 4.1 Point Correspondences S.14-16 - 4.1.1 Linear Method S.14-16 - 4.1.2 Non-linear method S.16 - 4.2 Line correspondences S.17-18 - 4.2.1 Linear method S.17 - 4.2.2 Non-linear method S.18 - 4.3 Experiments and results S.18 5 Camera pose estimation S.19-28 - 5.1 Description S.19-23 - 5.2 Pose by successive approximations S.24-25 - 5.3 Planar objects case S.25-28 6 Neck-eyes calibration S.29-47 - 6.1 Estimation of the HtoCi transformations S.31-32 - 6.2 Summary S.33 - 6.3 Estimation of the W toN transformation S.33-34 - 6.4 Experimental Results S.34-38 - 6.5 Stereo vision - Epipolar geometry S.39-47 - 6.5.1 Base line S.39-40 - 6.5.3 Epipolar lines S.40-41 - 6.5.4 Essential Matrix and Fundamental Matrix S.41-43 - 6.5.5 Stereo Matching Process S.43-44 - 6.5.6 3D reconstruction from 2D image points S.45-47 7 Arms-eyes calibration S.48-51 - 7.1 The \one eye looking" method S.48-50 - 7.2 Experimental results S.50-51 8 Conclusion and future extensions S.52-56 - A Rotation and Unit Quaternion S.52-56enkinematic calibrationcamera calibrationobject posestereo vision005006629report