Motion parameters for unmanned vehicle from an image sequence

Presents motion parameters estimation for an unmanned vehicle using a feature points extracted from a monocular sequence of images. The estimation assumes a fixed environment and a moving camera, mounted on the vehicle. The estimation process uses a recursive algorithm based on the extended Kalman filter which contains the dynamics of the vehicle. The simulation results in this work based on the X-RAE1 UMA model. It shows that the EKF estimator converges rapidly to the real values of motion parameters. A simple algorithm is also developed that avoids the correspondence and occlusion problems associated with feature tracking algorithms.