Global rotation estimation for stereo camera systems

This paper discussed the calibration of the relative pose of the stereo-camera. The rotation matrix estimation involved in this problem is typical instances of the single rotation averaging and conjugate rotation averaging problem. In order to avoid the ambiguity caused by the angle-axis and quaternion representations and obtain the global solution, a convex relation method is proposed for chordal metric under L-2 norm. In addition, a weighted robust cost function for the rotation averaging problems is also proposed to deal with the additional information about the measurements of the rotation and all the data including the outliers. The experimental results show that this method works efficiently and reliably. Meanwhile, the optimality is certified most of the time within a few sequence of linear matrix inequalities. (C) 2015 Elsevier GmbH. All rights reserved.