http://en.wikipedia.org/wiki/Photogrammetry

Photogrammetry refers, essentially, to measuring stuff using photographs. It's been used in traditional surveying for years (here's a paper from 1925-1926 on the subject http://www.iop.org/EJ/abstract/1475-4878/27/2/301). There seems to be a fair bit of commercial photogrammetry software out there. The main open source software I found was Ossim (pronounced Awesome!) (http://www.ossim.org) and it does seem to be designed with surveying in mind although I'm not quite clear what its capabilities are.

I've recently noted that there's a proposed building attribute (http://wiki.openstreetmap.org/index.php/Building_attributes) for height (amongst other interesting things). Measuring a building's height isn't easy to do accurately with a GPS (firstly because elevations don't seem to be very accurate on a basic GPS and secondly because you can't always get onto the roof!). Most of use don't carry a theodolite (http://en.wikipedia.org/wiki/Theodolite) to measure this sort of thing - even if we did, they're a bit unwieldy to tote around on a bike!

I've been pondering how feasible it might be to estimate approximate building heights using geotagged photographs and a much simpler piece of software. You would need (for a simple setup):
1) a reasonably level surface, lets assume you can be roughly at the building's ground level
2) photographs from a couple of different distances - at known positions, provided by the GPS
3) the position of the building, either by walking up to the building with your GPS or by looking it up on a satellite map (or OSM itself!)

It should then be possible to calculate the building's height using simple trigonometry, based on the GPS data and measurements from the photograph. I'd say it would be fairly trivial to write a little program that pops up a window with the photographs in and allows you to "draw" the height measurement you're interested in onto the building. It'd then do all the maths for you involving the length of those lines, the GPS co-ordinates, etc and crank out an estimate. It's waaaay off being a "proper" photogrammetry program but it might be useful for very quickly estimating the heights of buildings photographed during a mapping trip.

If collecting this sort of information could be made easy enough then rendering cities from OSM in 3D could start to get quite interesting. I'm not 100% convinced that the measurements would be accurate enough to be useful. Perhaps only a more elaborate solution (maybe look at Ossim for this!) would work, if it's practical at all.

I wonder if anyone else has thought about / tried this or has any comments?