An artificial intelligence enhanced star identification algorithm
- PDF / 857,788 Bytes
- 10 Pages / 595.276 x 841.89 pts (A4) Page_size
- 107 Downloads / 257 Views
1
Frontiers of Information Technology & Electronic Engineering www.jzus.zju.edu.cn; engineering.cae.cn; www.springerlink.com ISSN 2095-9184 (print); ISSN 2095-9230 (online) E-mail: [email protected]
An artificial intelligence enhanced star identification algorithm∗ Hao WANG†‡1 , Zhi-yuan WANG1 , Ben-dong WANG1 , Zhuo-qun YU1 , Zhong-he JIN1 , John L. CRASSIDIS2 1School
of Aeronautics and Astronautics, Zhejiang University, Hangzhou 310027, China
2Department
of Mechanical and Aerospace Engineering, University at Buffalo,
State University of New York, Amherst, NY 14260-4400, USA † E-mail:
[email protected]
Received Oct. 30, 2019; Revision accepted Apr. 14, 2020; Crosschecked July 20, 2020
Abstract: An artificial intelligence enhanced star identification algorithm is proposed for star trackers in lost-inspace mode. A convolutional neural network model based on Vgg16 is used in the artificial intelligence algorithm to classify star images. The training dataset is constructed to achieve the networks’ optimal performance. Simulation results show that the proposed algorithm is highly robust to many kinds of noise, including position noise, magnitude noise, false stars, and the tracker’s angular velocity. With a deep convolutional neural network, the identification accuracy is maintained at 96% despite noise and interruptions, which is a significant improvement to traditional pyramid and grid algorithms. Key words: Star tracker; Lost-in-space; Star identification; Convolutional neural network https://doi.org/10.1631/FITEE.1900590 CLC number: V447
1 Introduction As a kind of high-precision attitude determination instrument, star trackers are widely used in both orbiting and interplanetary spacecraft. Star trackers perform attitude determination by identifying stars in the field of view (FOV). Typically, a star tracker has two working modes, i.e., lost-in-space (LIS) mode and tracking mode. When prior knowledge of attitude information is unavailable, the star tracker operates in the LIS mode. In this case, a full-sky star identification algorithm (Spratling and Mortari, 2009) is required. Once the initial attitude has been determined, the star tracker switches to the tracking mode. In this case, the previously obtained infor‡ *
Corresponding author
Project supported by the National Natural Science Foundation of China (No. 6152403) ORCID: Hao WANG, https://orcid.org/0000-0002-0383-7258 c Zhejiang University and Springer-Verlag GmbH Germany, part of Springer Nature 2020
mation will be used to predict the current attitude so that the identification process is much easier than that in the LIS mode. Thus, a reliable LIS star identification algorithm becomes the major problem to be solved to obtain attitude information from a star tracker. Many star identification algorithms have been developed in the last 40 years to solve the full-sky star identification problem. The existing full-sky autonomous star identification algorithms can be divided roughly into two categories (Padgett and Kreutz-Delgado, 1997). The first kind tends to
Data Loading...
