Yes you can do it easily since WWII era,now with computers is easier, I suppose they will use brute force as usual for Google.

You could do it just taking infrared and ultraviolet photos, and comparing it with red, green and blue versions, leaves have their specific wavelength response, and some trees leave theirs on winter with specific yellows and browns. They could use it to segment for the tree specie.

Now for the three trunk position I suppose they will use their cars and bikes with LIDAR(very easy at ground level, think on a big kinect), or a very expensive plane with 3d recognition but this is problematic with trees that grow together and you don't know when one ends and the partner starts.

PS: I'm trying to detect trees position myself since a few years ago using aerial photography and ground cameras. I'm finishing a model of Casa de Campo: http://en.wikipedia.org/wiki/Casa_de_Campo

I imagine that it would be a very rough estimate. In areas of dense tree coverage in parks or backyards Google probably just plants trees to cover the area since it would be nearly impossible to actually differentiate each individual tree in a canopied area. Trees that are out in the open, standing alone can obviously be counted, but trees in groups probably can't.

That's assuming they're doing the tree counting with imagery. The newer Google Street View cars I've seen have LIDAR rigs on top, so they could be mapping trees that way.

But presumably they aren't driving through most of the forests, right?

I don't see any reason they couldn't do the same from the air. You can tell foliage with LIDAR, and I'd imagine you could get a pretty good approximation of trunk location pretty easily by looking at the foliage height.

Fair enough, but do they collect their own aerial data? I was under the impression it's all collected through by third parties. Also, do they use anything other than satellite imagery right now? Is LIDAR effective at that range?

Yeah, they have some aerial data. For example: zoom in on downtown San Jose in satellite view. The perspective view is aerial photography, and is copyrighted by Google.

Although, now that I've done more digging, it looks like they're using satellite imagery for at least some of this, along with LIDAR and infrared techniques: http://static.googleusercontent.com/external_content/untrust...

Just a hunch: I suspect they're not auto-detecting the trees, but rather relying on (amazing) tools like

http://www.urbanforestmap.org/

That map only San Francisco? I'm going with the LIDAR theory.