Computationally speaking, I just don't think 2-d 'ground level' photographs have enough exact inofrmation in them to extract and position in a 2d/3d top-down image. You would need topographical information you just would not be able to extract.

Even if you could use solar positioning and other 'rough' guesses to work out lattitude and longitude, then you are still left with hundreds or thousands of square miles in which to match you inacurate datapoints against.

Also, just in terms of computational complexity, even if you could get some good topographical data out, you are looking at one hell of a number crunch.

Davin L.