Highly robust star identification for SINS/CNS integrated navigation via CCPBM feature construction and Manhattan-distance-based hierarchical matching

Le résumé fourni par la source

In airborne strapdown inertial navigation system/celestial navigation system (SINS/CNS) integrated navigation, vibration, high dynamics, occlusion, and stray light can cause centroid shifts, false stars, and missing stars, posing challenges to robust star identification, whereas enhancing interference resistance may increase computational and storage costs. To address these issues, this study co-designs local feature construction and matching. A centroid-constrained principal-component basis method (CCPBM) combines second-order moments and centroid information to establish an oriented basis and directly encode neighboring-star positions, yielding a compact, magnitude-independent feature through a simple construction process. Single-frame multi-pattern redundancy mitigates local-pattern corruption caused by false and missing stars. A Manhattan-distance-based hierarchical matching framework further combines pattern-completeness-based principal-star prioritization with hierarchical candidate screening to balance reliability and online efficiency. Results show that, with n =2, the identification rates remain 99.58%, 99.99%, 98.58%, and 98.57% under independent disturbances of a 4-pixel positional noise standard deviation, 30 false stars, 10 missing stars, and 1-mag magnitude noise with missed detections, respectively. The mean per-frame identification time is 3.36 ms, and the storage requirement is below 400 KB, demonstrating a favorable balance between robustness and resource overhead.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.