Any progress? I'm working towards something similar, currently based on ZoneMinder (www.zoneminder.com) running the cameras. The latest version has a new feature described as follows:
"When motion is detected the centre of the region of motion is written to shared memory, where is can be used for tracking. This can now optionally be a simple median of the motion extents or (in blob mode) a weighted centre for better location of irregularly shaped events."
My plan is to use two cameras and a "mesh" of the terrain topography to triangulate positions.
I haven't actually experimented with this yet, as I haven't updated to the latest ZoneMinder version (1.22) and I'm still working on pan/tilt mounts for the cameras... and other things...