An enhanced centroiding approach for accurate star detection under lunar surface noise conditions

Abstract

Star sensors play a crucial role in space navigation, providing reliable attitude determination for spacecraft, which is especially important for future lunar missions such as the International Lunar Research Station (ILRS). Operating on the lunar surface, however, introduces unique challenges. Strong solar reflections from the lunar regolith, together with cosmic and solar radiation generate significant noise that can degrade image quality and disrupt the star detection as well as reducing the centroiding accuracy. This research aims to develop an enhanced centroiding algorithm to overcome the noises in the lunar environment as well as to evaluate the performance of the proposed algorithm. The methodology incorporates several preprocessing stages, including Point Spread Function (PSF) modelling to simulate optical blurring and median filtering for noise suppression. Star regions are then segmented using the global thresholding method, where the threshold value is defined as a function of the maximum image intensity. The centroid of each detected star is computed using an intensity-weighted Centre of Mass (COM) approach. The accuracy of the proposed method is validated by transforming pixel-based centroid coordinates into celestial coordinates through an affine transformation, followed by star identification using the Hipparcos star catalogue. Experimental evaluation was conducted to imitate the lunar surface environment under simulated noise including Poisson-Gaussian noise, salt-and-pepper noise, speckle noise, and sunlight glare. The results show that the proposed algorithm outperforms conventional methods such as COM, Gaussian Fitting, and Sieve Search Algorithm (SSA), achieving the lowest average RMSE of 1.218 pixels and Euclidean distance of 1.143 pixels. It also maintains a low False Detection Rate (FDR) of 6.716% and angular distance errors below 0.05°. These results demonstrate that the algorithm is robust under low SNR conditions, making it suitable for reliable star sensing in long-term lunar missions. Keywords: Star sensor, lunar surface navigation, centroiding algorithm, image preprocessing, attitude determination.