Star Identification using Neural Networks

Star trackers provide spacecraft with the most precise estimate of their orientation,or attitude, with respect to a fixed celestial coordinate system. The star tracker camera views a patch of the celestial sphere and attempts to recognize the stars contained within. Then from the known star positions it will calculate the attitude. A number of pattern recognition methods, each with various strengths and weaknesses, have been implemented in star trackers. The most challenging situation involves onboard autonomous identification. The limits on memory, power, weight, etc. place severe constraints on the processing available. We discuss here some neural algorithms and the kind of devices in which it might be implememented.