Some moons ago....
I was doing some work for a company that needed face recognition software written from scratch. Their language of choice was Microsofts Visual Basic, and whilst it's quite nice (especially since the advent of .net), it was not really the right tool for the job; however, I was but a mere programmer and the tools of my trade at that time were not my choice to make.

I basically started by chopping up pictures of faces into triangles and to cut a long and extremely boring and mathmatical explanation short, I used pythagorus to determine a recognition pattern and had an internal 'score'. If the 'score' was over a certain number, then in all likelyhood the person you were looking for was the person in the picure. This has given me an idea for a new application, but that's not for here.

At the time and at home, I had two cats.

I had trouble with other cats in the neighbourhood coming into my back porch and eating my cats food. Of course, I could simply have moved their bowls into the kitchen and let the (our) cats in and out when they needed feeding; but...that wouldn't have been anywhere near techie enough now would it?

At the firm I was doing the work for I had a colleague who was heavily into robotics. Indeed, his house would be the winner of this QoTW hands down, however, he gave me a voltage regulator and a serial controller at some point as I was making a 'pan and tilt' camera mount. The cat problem however, had given me a new use for the servos and donated equipment:

Yes. An electronic cat flap.

Now, this may seem misguided, but it worked...kind of.

I rigged up a large wheel that I made from ply as a gearing for the cat flap. I had a small carbon rod that was attached to the top of the gear wheel and the other end to a servo. The wheel would turn and in turn pull another rod (well two rods) over the top of a pivot and the cat flap would lift. Once motion was detected on the inside, the servo would receive a signal through the serial port and close the flap, just so I knew the cat was inside and I wasn't about to attempt to cut it in half.

An old (although it wasn't old at the time) Tosh laptop was the controller for this and I made a shelf for it in the porch connected wirelessly to a hub - remember this was long ago, wireless was not only expensive but far worse than it is now, but it worked most of the time.

I had two cameras at the time, as my ISP had a 'special' on the old logitec cam - 45 quid for one! You find them at boot fairs now for around 20p. Getting two idential webcams setup in Windows 95 was no mean feat I can tell you. Conflicts? Let me tell you about conflicts.....(no, I'll spare you)

I figured that if I could only get the cat to look directly into the web cam I could use the same software I'd written for the firm I was working for and get it to recognise my cats only. To do this, I had a little flashing LED next to the camera which started when the cam (or the program) detected motion, and thus attracting the cats attention.

I told you this was misguided.

Suprisingly, it did actually work, and I have an old handycam vid of it in practice somewhere. The only problem was that even if I could get the cat to look at the camera for long enough...all bloody cats look the same. Or at least they have the same symetrical features to their faces!

I took the bowls indoors in the end and simply let the cats in to eat.
(dchurch bah, Thu 20 Aug 2009, 21:35, 2 replies)

Perhaps
Factoring in their colour + markings would have bolstered your success rate?
(zero.coke, Thu 20 Aug 2009, 22:15, closed)

I
did think of doing that at the time, but to be honest, I had to cut my losses somewhere.

The main problem was getting the cat to look into the camera long enough to be recognised.
(dchurch bah, Fri 21 Aug 2009, 9:05, closed)