An End-to-End Identification Algorithm for Smearing Star Image

In the previous study, there were a few direct star identification (star-ID) algorithms for smearing image. An end-to-end star-ID algorithm is proposed in this article, to directly identify image from sensors with fast attitude maneuvering. Combined convolutional neural networks and self-attention mechanism of transformer encoder, can effectively classify star. Through feature extraction position encoding, learn stars generate semantic information realize The also solve problem low rate due long exposure time images. A dataset dynamic analyzed constructed based on multiple angular velocities. Experiment results show that, compared representative algorithms, improved at high When three-axis velocity 10°/s, still 60.4%. At same time, has good robustness noise magnitude noise.