If it were my lab, I would not try to use a still camera at all, but rather put the plate down on a programmable grid and aim a video camera at it. You don't need a microscope and if you are clever enough, you can write pattern recognition software to count for you.
I really have zero knowledge in photography, but I intend to learn.
I guess using a video camera would require a sliding mechanism for the plates to ensure a fixed rate of movement - or not necessarily?
Can you elaborate on the technical details?