Estimation and prediction for moving object pose

Position and orientation estimation of the object, which can be widely applied in the fields as robot navigation, electro-optic aiming system, etc, has an important value. The algorithm to determine the target's position and orientation with the image coordinates of feature points is very important in pose estimate technique. In this paper, a novel pose estimation and prediction method based on five coplanar reference points is presented. First according to the coordinates of the feature points in the world coordinate system and that on the CCD imaging plane, two linear systems could be established based on the perspective projection model and the quaternion transformation matrix of target is solved. Thus the position and orientation of the target is worked out. Considering the blind area between the two sample times, kalman filter theory is adopted to predict the pose of the moving object during the blind area, and obtain the optimal estimation of target pose at sample time. The application of kalman filter theory eliminates the measurement error induced by various interference factors effectively and provides advance motion information for subsequent tracking equipments, which finally fulfill the real-time request of tracking system.