Real-time pose estimation and tracking of rigid objects in 3D space using extended Kalman filter

Estimation of the orientation and location of a rigid object in 3D space using its 2D projection is considered in the current paper. In this case, a state-space model is proposed and applied to an Extended Kalman Filter (EKF) which estimates and tracks the parameters of 6 degree of freedom (DOF) movement (rotation and translation) of the object in 3D space. At the first stage, a matrix which projects the feature points in 3D space to 2D space is introduced based on the rotation-translation equations. Secondly, using the arrays of this matrix, the coordinates of the feature points in the image space are modeled by a state-space model. Finally, the 6DOF parameters are estimated using an EKF. The proposed algorithm is then evaluated by simulating the human head movements and the simulation results reveal that it can estimate the Euler angles with the estimation error less than 0.2 degree.