Robot Pose Estimation and Accuracy Analysis Based on Stereo Vision

摘要

In order to meet the stereo vision navigation of the autonomous robot, stereo vision initial pose estimation based on unit quaternion is put forward and pose estimation accuracy is predicted and analysed. The research provides basis for further improving the pose estimation accuracy of robot. Gauss-distributed error covariance matrix propagation model was adopted to deduce the error propagation formulas from 3D landmarks errors to final output pose errors of robot. The rotation covariance and the translation covariance are detached by the correlation analysis of rotation and translation of robot pose. Covariance estimation is realized effectively. Experiment shows realistic accuracy of pose estimation based on unit quaternion method and accuracy of error model deduction are the same basically. Accuracy analysis method of pose estimation in this paper is practicable.