You shouldn't need to fit a gaussian, instead compute the center of mass in a suitable area around and including the brightest pixel. I did just that for multi-star tracking, and it runs two of my automated observatories very well.