Stable vision-aided navigation for large-area augmented reality

摘要: In this paper, we present a unified approach for drift-free and jitter-reduced vision-aided navigation system. This is based on an error-state Kalman filter algorithm using both relative (local) measurements obtained from image motion estimation through visual odometry, global as result of landmark matching pre-built database. To improve the accuracy in pose augmented reality applications, capture 3D local reconstruction uncertainty each point covariance matrix implicity rely more closer points filter. We conduct number experiments aimed at evaluating different aspects our framework, show can provide highly-accurate stable indoors outdoors over large areas. The results demonstrate long term stability overall intended to solution camera tracking problem applications.