Relative Attitude Estimation for a Uniform Motion and Slowly Rotating Noncooperative Spacecraft

Abstract

This paper presents a novel relative attitude estimation approach for a uniform motion and slowly rotating noncooperative spacecraft. It is assumed that the uniform motion and slowly rotating noncooperative chief spacecraft is in failure or out of control and there is no a priori rotation rate information. We utilize a very fast binary descriptor based on binary robust independent elementary features (BRIEF) to obtain the features of the target, which are rotational invariance and resistance to noise. And then, we propose a novel combination of single candidate random sample consensus (RANSAC) with extended Kalman filter (EKF) that makes use of the available prior probabilistic information from the EKF in the RANSAC model hypothesis stage. The advantage of this combination obviously reduces the sample size to only one, which results in large computational savings without the loss of accuracy. Experimental results from real image sequence of a real model target show that the relative angular error is about 3.5% and the mean angular velocity error is about 0.1 deg/s.