Estimation of The Location and Orientation of a Camera in Short Image Sequence Using Extended Kalman Filter

Estimating the camera location and orientation from an image sequence is wildly used in many applications such as Augmented Reality and Robot Navigation. In this paper a new hybrid method is proposed. For this purpose, we apply Extended Kalman Filter to estimate the camera trajectory with 6 degree of freedom. The main difference between the proposed method and other filter-based methods is that in the proposed method, 3D points of the world model have been removed from the state vector and alternatively, Multiple View Geometry methods have been used to initialize the world model without uncertainty. As a result, the complexity of the algorithm, which is one of the shortcomings of the filter-based methods, has been reduced. Computer simulations and experimental results show that the proposed method improves pose estimation compared to PnP methods.