Estimation Of Relative State Between Noncooperative Spacecraft Based On Stereovision

For the advanced operating standard in space missions such as on-orbit servicing and debris removal, research on effective and autonomous relative navigation applied in space proximity operations has become a focus problem recently. This thesis does research on the relative navigation adopting binocular stereovision measure in autonomous rendezvous of non-cooperative spacecraft. Processing the image information with least-squares, reconstruction of the relative state is obtained. Precision of the estimation is then enhanced by using filtering based on model of the relative pose and motion. Also an estimation of relative state associated with inertia uncertainties is implemented in this thesis.A six-DOF formulation of relative motion is built by combining models of the relative position and the relative attitude, for the reason that the vision-based measure can only sense surface features of the target. Considering the random distribution of feature points on non-cooperative spacecraft, construct a model of non-center of mass points between the chaser and the target spacecraft. It is shown that, the deviations in relative position induced by kinematic coupling of non-center of mass points obviously have effect on the space proximity operations.Based on the study of pose transformation and imaging principle in computer vision, the imaging characteristic for targets in space is analyzed. On the account of possible fuzzy and degradation on the image, study on image processing technologies such as noise reduction, restoration and enhancement has been made. To verifying the effectiveness of processing algorithmic, some classic images are adopted in demonstration.A binocular visual observation model is established by using computer vision theory. According to the image information of a single frame-pair, combining with image velocity and stereo-vision disparity, the relative state is reconstructed by using least squares. To improve the estimation of relative state, leading in linearized model of the relative motion, reconstructions are processed as pseudo-measurements for an EKF scheme, which is utilized to sequentially copy with continuous image frames. The results of compared simulations indicate that filtering makes obviously improvement in the estimation of relative state for the single-frame-based reconstruction.Regarding to the problem of inertia uncertainty in relative navigation for non-cooperative spacecraft, state reconstruction and sequential estimation are used in the context of small amount of a priori inertia information. A scheme consists of several concurrently working EKFs is designed for estimating the relative on the account of some inertia hypotheses. Using a Bayesian maximum posterior estimation, the hypothesis nearest to the real inertia tensor is identified by comparing accumulated residual of each EKF, and the corresponding estimation of relative state is naturally the best. Results of simulations demonstrate that, solutions proposed here can effectively improve the estimation of relative position and attitude in the case of inertia uncertainties by making sufficiently use of a priori information and measurements.