A rather simple improvement would be to have all the lights flash their binary encoded position for the calibration step. You go down from calibration time O(n) to O(ln(n)). Using red/blue would be better than using on off for this since you can get a more accurate position read by averaging ln(n) shots for the location of a light.
I believe something similar is used by some projection mapping software.