Star trackers are essential for absolute attitude determination in space using high precision centroid calculations on stars detected in images. This study presents a CNN-based method that significantly enhances the performance of star detection and centroid calculations under high sensor noise and stray light conditions.

• Advanced CNN techniques improve star detection and reduce errors from sensor flaws.
• Real-time processing capabilities on low-power AI processors enable on-the-fly computations.
• Extensive tests confirm reliability over existing methods in various conditions.

This innovative approach’s real-time processing capabilities and robustness against severe disturbances make it paramount for future space missions, potentially enhancing the accuracy and reliability of autonomous spacecraft navigation.