A star identification algorithm based on 1D Conv Neural Network

Abstract A novel star identification algorithm based on 1D-CNN for celestial navigation is introduced. This approach constructs a distinctive star pattern using radial and angular separations and cyclic angles. Several tests have been conducted on simulated star images, demonstrating the proposed algorithm’s superior performance over conventional algorithms and high resilience to three error types, including spurious stars, positional deviations, and magnitude uncertainties.