Relative Stereovision-Based Navigation for Noncooperative Spacecraft via Feature Extraction

Spacecraft proximity operations play an important role in guaranteeing the capture success of noncooperative targets. This article technically presents a relative stereovision-based navigation scheme that ensures accurate relative pose measurement as well as practical motion estimation for proximity operations. Specifically, features parameters of the line and circle on the noncooperative target are first obtained by the different image processing extraction algorithms. Then, the center position and normal of the circle are derived by orthogonal transformation from the circle feature parameters, and two features descriptors are defined by the redundant parameters in the stereovision to evaluate error levels of the extraction algorithms. In particular, the unavailable roll angle around the circle normal is recovered by incorporating the line and the descriptors by utilizing the least squares method, wherein the relative attitude is proved less affected by the features extraction errors, and thus, the measurement accuracy is theoretically improved. Furthermore, considering the Mahalanobis distances of the attitude and position, two outlier detectors are coupled with the 3
$\sigma$
rule in a multiplicative extended Kalman filter to achieve outlier rejection and motion estimation, respectively. Finally, numerical simulation and physical experiment are given to validate the effectiveness of the proposed approach.