Prediction and estimation of pose accuracy of two-ocular optical tracker

This paper investigates an approach of predicting and estimating pose accuracy of two-ocular optical tracker in augmented reality systems. Gauss-distributed error covariance matrix propagation model was adopted to deduce the error propagation formulas from camera image errors to final output pose errors of the tracker, and the variation characteristics of the tracker pose error were analyzed through experiments under static and dynamic tracking states. Experimental results are consistent with the predicted data, which confirms the dominant effect of camera extrinsic parameter errors on tracker pose accuracy and validates the usefulness of our proposed approach to predicting and estimating pose accuracy.