Research On Pose Estimation Of Moving Rigid Body Based On Kalman Filters

Camera-based pose estimation problem is a classical problem in computer vision,due to its application values in rigid body pose estimation,it still attracts increasing research attention in the last decade.It has been widely used in aerospace,visual servo,robot navigation and other fields.In this thesis,the Kalman-based pose of rigid body pose estimation is investigated by applying the Kalman filter approach,aiming at improving the accuracy and robustness of the vision pose estimation system,and achieving the accurate tracking of the target.The main work and results of the thesis are as follows:1.The camera imaging model and representation methods of pose are introduced,and the relationship between the coordinate transform is analyzed.An image processing method is proposed to extract the mark.Finally,the experiment platform including the software of the pose estimation system is designed.2.For single camera pose estimation system,the nonlinear Kalman filter is applied,including the adaptive extended Kalman filter,and the unscented Kalman filter.A number of simulations and experiments are presented to verify the effectiveness of the proposed methods.The simulations and experiments show that the proposed methods provide higher estimation precision as compared with existing methods.3.For multi-cameras pose estimation system,the unscented Kalman filter is designed as the local filter,two kinds of distributed fusion methods are designed,namely,the distributed fusion method with matrix weights and the CI fusion method.Both simulations and experiments are presented to verify the effectiveness of the proposed methods.4.For the condition of occlusion in pose estimation,the feasibility of distributed fusion methods are analyzed.To obtain satisfactory estimation accuracy in the case of occlusion,the existing distributed fusion method is improved,and the effectiveness of the improved method is verified by experiments.Finally,the main results are summarized,and some prospects of further research are analyzed.