Height data is incredibly sparse for OSM on average. I’ve been trying to think of a good way to visualize buildings without adequate height data… my goal is to present a reasonable appearance with a 3d view, even if the height data is grossly inaccurate. Just using a random value looks quite poor overall – especially in suburbs and areas that don’t have many high buildings. Another method that might work out better would be to use the area profile as a basis for estimating height. Does anyone have any tips or references for deriving an estimated building height from its profile? Again – I don’t care too much on accuracy here, but if I’m viewing a residential street with a bunch of detached houses, I shouldn’t see buildings as tall as sky scrapers
Why not use the buildings which do have a height tag as your guide? First you could calculate a height / area ratio (or even a distribution as it’s unlikely to be linear) for a wide area - could even be the whole world / country. If in your view, there are no buildings with height, then use that as your guide. If in your viewing area there are some buildings with height data, you could use those. This means that in a residential street, your guesses will be mostly 1-2 floors, whereas in a business park / Manhattan, they will be higher.
Building type and area type are further guides. In residential areas, most untagged buildings will be houses. Churches are generally tall buildings, etc…
Thanks for the detailed reply. This would basically be creating valid building data for the map (which I may try to do at some point regardless), but doesn’t let me create building height data on the fly.
I was hoping there was a paper or something which took a look at this, and had certain metrics (for instance, a building with a sizeable hole in it tends to be only a few floors, even if its total area is quite large; same goes for building profiles that have an extreme height to width ratio – so shape, along with area plays a role).
Creating a distribution is a pretty cool suggestion… I like the idea of being able to throw an area profile into a function and have it return an estimated height. Osmium might be a good tool for this.
Also side note/question: The wiki says we’re allowed to use Bing aerial imagery to extract OSM data (barring roads and labels). Does this include their bird’s eye (isometric) views? I feel like you could automate and extract tons of useful building data with the right image processing techniques. Example view: http://binged.it/IrN4I5
Ther is no connection between use and estimated height. The autors wrote (page 9) e.g.: “Residential buildings are predominantly one or two story structures, with a few high-rise buildings (> 15 m) present (likely apartment buildings or condominiums)” .
Maybe in Houston.
I teached 17 years long on the University in Hamburg ( city planning anc civil engineering). Of course, You can find a kind of correlations for the citie, but thea are different. We did with students analyses for some cities in Germany. They are different. Generalization is not really helpful. I can understand Your goal, but it doesn´t work. Of course, Yo can see some 3D results but they are far away from the reality.