NESP

Neural Estimation of Spacecraft Pose.

If first html tag is indented, and this include is called after a list, the html tag is considered a list element, and things go wrong. Leaving a hidden unindented line here
Period of Performance: 2022–2026
Role: KU Leuven PI
Additional Contributors: Antoine Legrand, Christophe De Vleeschouwer.
sponsor Funded by Service public de Wallonie (SPW).
sponsor Funded by Aerospacelab.
project thumbnail

Neural Estimation of Spacecraft Pose.

If first html tag is indented, and this include is called after a list, the html tag is considered a list element, and things go wrong. Leaving a hidden unindented line here
Period of Performance: 2022–2026
Role: KU Leuven PI
Additional Contributors: Antoine Legrand, Christophe De Vleeschouwer.
sponsor Funded by Service public de Wallonie (SPW).
sponsor Funded by Aerospacelab.
project thumbnail

We study the problem of estimating the pose of a known spacecraft in orbital lighting conditions. Building on the state of art in terrestrial application, we specifically focus on a model that solves the problem in an end-to-end fashion, which delivers high accuracy at low computational cost, thanks to parallelizability.

If first html tag is indented, and this include is called after a list, the html tag is considered a list element, and things go wrong. Leaving a hidden unindented line here
Overview of the existing methods to estimate a spacecraft’s pose. (a) Direct methods use a convolutional neural network to directly predict the target 6D pose. (b) Keypoints-based methods exploit a convolutional neural network to predict pre-defined keypoint coordinates which are then used to recover the pose through a Perspective-n-Point (PnP) solver. (c) Our solution combines both methods as it relies on a first, convolutional, neural network to extract the coordinates of pre-defined keypoints and a neural network to predict pose from those coordinates. The solution is therefore both keypoints-based and end-to-end trainable.

State-of-the-art methods for estimating the pose of spacecrafts in Earth-orbit images rely on a convolutional neural network either to directly regress the spacecraft’s 6D pose parameters, or to localize pre-defined keypoints that are then used to compute pose through a Perspective-n-Point solver. In our first paper (Legrand et al., 2023), we study an alternative solution that uses a convolutional network to predict keypoint locations, which are in turn used by a second network to infer the spacecraft’s 6D pose. This formulation retains the performance advantages of keypoint-based methods, while affording end-to-end training and faster processing. Our paper is the first to evaluate the applicability of such a method to the space domain. On the SPEED dataset, our approach achieves a mean rotation error of 4.69° and a mean translation error of 1.59% with a throughput of 31 fps. We show that computational complexity can be reduced at the cost of a minor loss in accuracy.

References

  1. Legrand-2023-AI4Space-0.jpg
    End-to-end Neural Estimation of Spacecraft Pose with Intermediate Detection of Keypoints.
    Antoine Legrand, Renaud Detry, and Christophe De Vleeschouwer.
    In Computer Vision – ECCV 2022 Workshops, 2023 (Best presentation award).
    Bib DOI PDF