<![CDATA[The need to determine the orientation while in "Lost-In-Space (LIS)" is essential for spacecraft navigation. Star pattern recognition, also known as the star identification algorithm, plays a vital role for a spacecraft in LIS mode. Data-driven solutions for this type of problem are becoming more captivating due to their stochastic nature. This paper presents an efficient feature extraction method for the LIS star identification algorithm using a convolutional neural network. The net pattern and the multi-triangles feature extraction methods are implemented on the model. The proposed idea is tested on several simulated star images having a field of view of 25 by 16 degrees. The obtained results show an improvement in the successful identification rate of star image classes. Furthermore, the algorithm shows promising running time and requires less onboard memory since it eliminates storing a star catalogue for the matching process.]]>
Copyrights © 2024
