Incremental validation test campaign of a CNN-based Image Processing algorithm for autonomous visual based navigation applied to the Hera mission

Tools

Kaluthantrige, Aurelio and Feng, Jinglang and Di Domenico, Gianfranco and Gil-Fernández, Jesús (2026) Incremental validation test campaign of a CNN-based Image Processing algorithm for autonomous visual based navigation applied to the Hera mission. Advances in Space Research. ISSN 0273-1177 (https://doi.org/10.1016/j.asr.2026.04.105)

[thumbnail of Kaluthantrige-etal-ASR-2026-Incremental-validation-test-campaign-of-a-CNN-based-Image-Processing-algorithm]
Preview
 Text. Filename: Kaluthantrige-etal-ASR-2026-Incremental-validation-test-campaign-of-a-CNN-based-Image-Processing-algorithm.pdf
License: Creative Commons Attribution 4.0 logo
Download (5MB)| Preview

Abstract

This paper presents the incremental validation test campaign of a Convolutional Neural Network-based Image Processing algorithm using as case scenario the proximity operations of the European Space Agency’s Hera mission around the target binary asteroid system (65803) Didymos. The algorithm estimates the position of the Center of Mass of both the primary and the secondary (if visible), the range from the primary, and the associated covariance matrices using the synthetic images generated with the SW Planet and Asteroid Natural Scene Generation Utility. The first step of this incremental validation process is an Open-Loop and Closed-Loop Model-In-The-Loop test, which uses the Functional Engineering Simulator of the Guidance, Navigation and Control system of the Hera mission. The second step is a bare-metal Open-Loop test run on the processing system of a Zynq 7000 System-on-a-Chip, considered in this work as the representative of a typical spacecraft On-Board Computer. The third and final step is an Open-Loop Hardware-In-The-Loop test, which includes the Hera mission’s camera functional model and it is run at the GMV Optical Laboratory. The test case scenarios are the Early Characterization Phase, the Detailed Characterization Phase and the Close Observation Phase of the proximity operations of the Hera mission, when the vision-based navigation system is based on the centroid of Didymos. Therefore, the incremental validation test campaign presented in this work considers only the position of the Center of Mass of the primary estimated by the developed Image Processing algorithm. The results show that the algorithm is able to perform accurately its estimations across the multiple tests, with a slight dependency on the calibration of the camera. Nevertheless, the results show that prior to the implementation on a spacecraft On-Board computer the algorithm requires a dedicated memory optimization process.

ORCID iDs

Kaluthantrige, Aurelio, Feng, Jinglang ORCID logoORCID: https://orcid.org/0000-0003-0376-886X, Di Domenico, Gianfranco and Gil-Fernández, Jesús;
CORE (COnnecting REpositories)