[Joel J]

It seems to just average the centroids of both targets (so if it finds two lights, then following the centroid sent back would have you heading towards one of the "un lighted" columns). If the pixel count indicates the presence of two targets, you would possibly have to rotate the camera to home in on only one, then track again.

edit: I'm kinda curious how this announced feature works, as well. I shouldn't speculate based only on what I've seen.