Shape Reconstruction and Rotation Axis Estimation of Small Bodies Based on Structure-from-Motion

Abstract. Shape reconstruction and rotation axis estimation of small bodies essential for both engineering applications and scientific investigations. This paper presents a Structure-from-Motion-based (SFM) method for small body shape reconstruction and rotation axis estimation. The method enables fast and autonomous estimation of shape and rotation axis at relatively large distances during the approach phase with a limited number of images. Using near-hovering observations, sequential image poses are estimated via incremental SFM. The normal vector of the plane where the fitted space circle is located is identified as the small body’s rotation axis and transformed into the small body-centered J2000 inertial coordinate system. A global shape model is then generated through dense stereo matching. The proposed method is evaluated using both simulated and real mission data. A total of 75 simulation cases are designed, accounting for sun phase angle, approach angle, small body shape, and image count per rotation period. Results show that over 95% of cases achieve a rotation axis estimation error below 5°. When tens of images are captured per rotation period, the rotation axis can be estimated within minutes. Validation with OSIRIS-REx mission data for Bennu yields a rotation axis estimation error of approximately 1°, while dense reconstruction shows an average deviation of 2.55 m compared to the SPC shape model. These findings demonstrate the method’s effectiveness and suitability for small body exploration.