I'm drawing a line under ARM processors for now as the funding is here and moving forward with the working Vivotek device and independent temp/humidity loggers. The massive tuffbox full of lead acid batteries is unattractive, so now Plan A is in the bag with off the shelf parts and a correspondingly short lead time. I'm looking at another option that as been on the back burner.
Time lapse recording in the trail cameras I've looked at (bushnell / reconyx) is limited to 1 frame a minute. Looking at work in Canada and mean amphibian velocities (Hels 2001) it is know that this frequency is too slow, certainly for Rana temporaria! The Bushnell I have would also be a really shit choice of camera to put in a tunnel.
A system that can capture images every 10/20/30 seconds is going to need a bit more work;
I like the image of time lapse with PIC, (for battery consumption), but as it's already been done with Arduino it would seem sensible to go with what's out there. Either way it just seems like a matter of physically switching the shutter button; http://www.instructables.com/id/Time-Lapse-Photography/
Stripping the IR cut filter is another well trodden path; http://www.nigelrichards.org.uk/Experiments%20in%20Infrared.htm
Getting the images out of the camera while leaving it in situ relies on it behaving like a USB storage device - and no settings would need to be adjusted - ie 'PC mode'. maybe the intervalometer needs to be disabled during this task.
Then tie all this in with an SD card and 1-wire libraries... see earlier post.
Camera requirements that I can think of...silent(ish) and no focussing lights
wakes up ready to take photos with last settings.
takes a 32gb SD card
behaves as USB storage device
wide angle lens - or will take adaptor
decent sensor - low noise in low light
Reviewing the time-lapse images doesn't seem so bad once they are compiled as a movie. I've had some success running a time-lapse movie through motion detection too, (which is why the prototype bufocam is plugged into my media player). The smart solution will be something like a simplecv script I guess - for now watching it through seems ok.
All i need now is an Arduino and a compact camera to hack apart. Any donors out there please get in touch....