False star filtering and camera motion estimation via density-based clustering

Erdem Onur Ozyurt*, Alim Rustem Aslan

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

Star sensors serve as sophisticated instruments for determining spacecraft attitude, offering high accuracy to meet complicated scientific demands. However, their accuracy can be compromised by various sources of noise in the captured image, such as false stars arising from reflective objects or solar flares. Filtering out these false stars is essential for enhancing accuracy and reducing computational complexity. This study introduces an algorithm designed to identify and filter false stars while also estimating camera motion parameters, thus improving attitude determination performance. The algorithm operates by detecting isomorphic feature vectors via density-based clustering that is employed to discern false stars. Moreover, the variance in slope angles of true star pairs facilitates the derivation of an affine transformation matrix through a maximum likelihood estimator. It is a standalone algorithm that can be integrated into any star identification method to increase robustness to false stars while providing motion parameters to be used in recursive star identification algorithms to reduce complexity. The algorithm's effectiveness is evaluated through experiments on 1000 pairs of time-sequential simulated star images, in which the sensor parameters are taken from the SharjahSat-1 project, while also taking position and brightness noise effects into account.

Original languageEnglish
JournalAdvances in Space Research
DOIs
Publication statusAccepted/In press - 2025

Bibliographical note

Publisher Copyright:
© 2024 COSPAR

Keywords

  • Attitude determination
  • Camera motion estimation
  • Density-based clustering
  • False star filtering
  • Star identification
  • Star sensor

Fingerprint

Dive into the research topics of 'False star filtering and camera motion estimation via density-based clustering'. Together they form a unique fingerprint.

Cite this