This is the authors' abstract. We don't add key points for this paper.
Star-based navigation on the lunar surface is severely degraded by strong regolith reflections, abrupt illumination transitions, and sensor-induced noise, all of which lower the signal-to-noise ratio (SNR) and impair centroiding accuracy. Existing methods address this trade-off poorly: classical center-of-mass (COM) and Gaussian fitting are computationally light but noise-sensitive, whereas iterative weighting and learning-based approaches improve accuracy at the cost of high computational load and large training data. This study addresses this gap by proposing a low-complexity yet robust star centroiding algorithm tailored for lunar-surface imagery. The pipeline integrates median filtering for impulse-noise suppression, an adaptive global threshold (set at 30% of peak intensity) for star-region segmentation, and an intensity-weighted COM estimator for sub-pixel localization. The method was implemented in MATLAB R2023b and evaluated on 30 Stellarium-derived star fields, each corrupted by Gaussian, Poisson, salt-and-pepper, speckle, and solar-glare noise spanning SNRs from ?4.71 dB to 0.72 dB. Benchmarked against standard COM, Gaussian fitting, and the Sieve Search Algorithm (SSA), the proposed method achieves the lowest average root-mean-square error (RMSE = 1.218 pixels), the lowest Euclidean distance error (1.143 pixels), and the lowest false detection rate (FDR = 6.716%), corresponding to relative reductions of 21.5%, 16.7%, and 32.7% over the best baseline, respectively. The algorithm’s favorable accuracy–complexity trade-off makes it well-suited for resource-constrained onboard processors in future lunar exploration and autonomous spacecraft navigation missions. ABSTRAK: Navigasi berasaskan bintang di permukaan bulan terjejas dengan ketara akibat pantulan regolit yang kuat, perubahan pencahayaan yang mendadak, dan hingar daripada penderia, yang kesemuanya menurunkan nisbah isyarat kepada hingar (SNR) dan menjejaskan ketepatan penganggaran sentroid. Kaedah sedia ada gagal mengimbangi keseimbangan ini: kaedah pusat jisim (COM) klasik dan pemadanan Gaussian adalah ringan dari segi pengiraan tetapi sensitif terhadap hingar, manakala kaedah pemberat berulang dan berasaskan pembelajaran mesin meningkatkan ketepatan tetapi memerlukan beban pengiraan tinggi dan data latihan yang besar. Kajian ini mengisi jurang tersebut dengan mencadangkan satu algoritma penganggaran sentroid bintang yang teguh dan berkompleksiti rendah, khusus untuk imej permukaan bulan. Saluran pemprosesan menggabungkan penapisan median untuk menindas hingar impuls, ambang global mudah suai (ditetapkan pada 30% daripada keamatan puncak) untuk segmentasi kawasan bintang, dan penganggar COM berwajaran keamatan untuk penyetempatan subpiksel. Algoritma dilaksanakan dalam MATLAB R2023b dan dinilai ke atas 30 medan bintang Stellarium yang dicemari oleh hingar Gaussian, Poisson, garam-dan-lada, bintik, dan silau solar dengan SNR dari ?4.71 dB hingga 0.72 dB. Berbanding COM standard, pemadanan Gaussian, dan Algoritma Carian Ayak (SSA), kaedah yang dicadangkan mencapai purata RMSE terendah (1.218 piksel), jarak Euclidean terendah (1.143 piksel), dan kadar pengesanan palsu terendah (FDR = 6.716%), masing-masing mewakili pengurangan relatif 21.5%, 16.7%, dan 32.7% berbanding garis dasar terbaik. Imbangan ketepatan-kompleksiti yang baik menjadikan algoritma ini sesuai untuk pemproses dalam-pesawat dengan sumber terhad bagi misi penerokaan bulan dan navigasi pesawat angkasa autonomi pada masa hadapan.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Razaman et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: