Vision-Based Estimation of Small Body Rotational State during the Approach Phase

TL;DR

This paper presents a vision-based algorithm to estimate the rotational state of small celestial bodies during the approach phase, enabling autonomous characterization before orbit insertion.

Contribution

It introduces a novel method for estimating small body rotation axes and centers using feature tracking in images, tested extensively on synthetic asteroid data.

Findings

80% of test cases had rotation axis error below 10 degrees

Algorithm successfully estimates rotation axis and center in diverse lighting and orientation conditions

Validated on over 800 synthetic scenarios with Bennu and Itokawa asteroids

Abstract

The heterogeneity of the small body population complicates the prediction of small body properties before the spacecraft's arrival. In the context of autonomous small body exploration, it is crucial to develop algorithms that estimate the small body characteristics before orbit insertion and close proximity operations. This paper develops a vision-based estimation of the small-body rotational state (i.e., the center of rotation and rotation axis direction) during the approach phase. In this mission phase, the spacecraft observes the rotating celestial body and tracks features in images. As feature tracks are the projection of the landmarks' circular movement, the possible rotation axes are computed. Then, the rotation axis solution is chosen among the possible candidates by exploiting feature motion and a heuristic approach. Finally, the center of rotation is estimated from the center of brightness. The algorithm is tested on more than 800 test cases with two different asteroids (i.e., Bennu and Itokawa), three different lighting conditions, and more than 100 different rotation axis orientations. Each test case is composed of about 250 synthetic images of the asteroid which are used to track features and determine the rotational state. Results show that the error between the true rotation axis and its estimation is below $10^{\circ}$ for $80\%$ of the considered test cases, implying that the proposed algorithm is a suitable method for autonomous small body characterization.