Trying to determine the relative position of a person in a room

Wolfbrother

Hi,

I'm trying to control a simple output based on detecting where a person moves to in a room. It could be multiple persons and they can be separate. I'd like to single one person·out and follow them. My idea is to install a ccd/cmos imager in the ceiling and look for different color blobs. If I have a 12'x8' room, ideally each one foot square could be a pixel so I would only keep track of 96 pixels. The floor is light grey or white, so I am thinking people will stand out against that background. My questions are, does anyone know a sensor they would recommend for this, is the basic stamp fast enough to process this much information and does anyone have any experience trying something like this? Lastly if someone else has a better way to determine people's position in a room, I'm all ears.

Thanks.

SRLM

Parallax has their CMU cam, and there is code to make it track colors when mounted on a BOE-BOT. One problem that I can think of is that your ceiling will have to be pretty high if you want the tracking to be accurate and avoid using a wide angle lens.

Wolfbrother

Hi,

With the CMU cam it's 80x143 and I can just ignore values outside of my mask, but I was hoping someone would know of a sensor with ( yes I know this is strange in modern times) less resolution, the less information I have, might actuall be better... On the CMUCam,I was more thinking that it seems to be optimized for close in viewing and if anyone had any experience with greater distances. I should have put in my first post that the ceiling is about 14 feet high. I don't mind the fish-eye effect at the edges of the room, since people tend to not gather in corners either. The tracking just needs to close, not accurate. I could have done a matrix of photo sensors in the wall, if I could figure out how to select the same person if two people crossed, I just thought this might be easier. Thanks for your thoughts.