marlor:
Thanks for the help, this is supposed to be a prototype like thing of sensors for veldt fires, so thats why smoke wouldnt work, and why we speculated about IR or UV
Because you mentioned school I put two and two together and came up with five thinking it was an indoor problem.
The wiki link still applies. Outdoors sunlight is going to be a big issue. You also need to look at sensor costs, they might be expensive. Range might also be important. I can see the visible light from a fire from a long distance away but I can only feel the heat on my skin when I am quite close so I am not sure how far IR can be 'seen' from.
The ideas in the wiki about comparing the ratios of different wavelengths might be worth exploring. IT also might be worth thinking about taking still images and subtracting one from the other to see what has changed. Fire and smoke might cause big changes (but the Arduino is not good at image processing).