Here's a tricky one. Count the number of people in the room.

If I had an infrared photograph I could count the number of bright areas in the image to get the number of occupants (or at least a statistically close value). Assuming I was able to extract the chroma value of each pixel and insert to a record, (plus a position field), I would have a map of the image.

I have a vague idea about creating a virtual zone of say 100 pixels (10x10) in the top left of the image and doing a search for the records in the zone. I'd then need to find the record with the highest value. This would then form the centre of my next search and using get nth record re-evaluate my new search zone; which would continue until the result was the centre. The zone would then move across to the next adjacent zone and repeat.

Any thoughts or better ideas?

This is not really an issue for a Filemaker forum but is, perhaps for that reason, more interesting. I took a quick look and, by adding a programming language to my search criteria, came up with this:

http://pythonvision.org/basic-tutorial

A quick scan shows it deals with a problem very similar to yours - finding nuclei in a sample. It also uses very high level functions making the programming look relatively easy.

Python is callable from Filemaker, is installed on OS X and free to install under Windows.

No idea what it means but seems to get the result - thanks