Title: Design and implementation of attitude determination system for Cubesat using image processing with deep learning
By Alireza Amirkhani
Supervisor: Dr. Mehran Mirshams
Advisors: Mr. Mohsen Shahmirzaei, Ms. Niki Sajjad
Chairmen: Dr. Alireza Basohbat Novinzadeh (K.N. Toosi University of Technology), Dr. Hamed Alisadeghi (K.N. Toosi University of Technology)
October 17, 2021.
Abstract: For many years, satellite attitude determination has been done using sensors such as sun sensors, earth horizon sensors, magnetometers, and star trackers, each of which has advantages and disadvantages in terms of accuracy and cost. As a new method to increase the accuracy and reduce the cost of the satellite sensor package, using artificial intelligence technology and in particular, the machine learning algorithms for satellite attitude determination have been proposed.
This research presents an algorithm based on deep learning for attitude determination, considering many applications of computer vision in improving the accuracy of imaging sensors. In fact, the proposed algorithm is extracted using a deep neural network and been tested on the images captured by the simulated camera sensor as a satellite payload. Earth images were also generated in this research using simulations in STK software. The algorithm is implemented in the laboratory and on the actual hardware. This method is effective in reducing the size, mass, and cost of the attitude sensor while increasing the accuracy of the attitude determination system.