[OSM-dev] OSM Level of Detail for GSoC 2017

Balázs Dukai balazs.dukai at hotmail.com
Fri Dec 16 13:18:23 UTC 2016


Hello everyone,

as I'm a newcomer to this list please let me introduce myself. My name 
is Balázs Dukai, I'm currently studying geomatics engineering at TU 
Delft, Netherlands and I was not active in the OSM development community 
until now. I'm moderately experienced with OSM, last year I organized 
two Missing Maps parties at TU Delft and I use OSM mainly for 
downloading vector data to prepare base maps. This summer (2016) I 
participated in the GSoC, however under the umbrella of the R project 
and I developed a package to manage animal trajectories in PostGIS and 
R[1,2]. Since then I started working on my thesis which brings me to 
this mailing list.

In my research I'm developing a method to automatically infer the Level 
of Detail (as in CityGML) of heterogeneous 3D city models. Experience 
shows that several 3D city models contain buildings that are tagged as 
e.g. LOD2, however they are modeled as LOD1. This heterogeneous data 
negatively influences the accuracy of analysis that depends on the 3D 
geometry, such as energy profiling, solar irradiation, wind flow 
analysis. Also it is good to know that the data set one purchased is 
truly in-line with its description, e.g. all buildings are really LOD2. 
Visually assessing the LOD of each building is not feasible at a city 
scale, hence the automatic inference.

Now, in order to have an idea about the LOD of the 3D building, a good 
way to start is to evaluate its footprint. As the geometric detail of 
the footprint indicates the geometric detail of the building. My 
assumption is that having an indication (a tag maybe) of the detailed 
LOD[3] of the 2D building footprints in OpenStreetMap could provide some 
benefits, such as

+ identify areas where the buildings are mapped in a low LOD and 
could/should be improved,

+ users could easily see the LOD of buildings (LOD tag) when 
programmatically querying/downloading OSM data and better evaluate its 
fitness-for-purpose.

Therefore I propose that in the scope of GSoC 2017 I work towards a 
tool/method that automatically appends the detailed LOD[3] tag to 
building footprints in OSM. Maybe even the potential LOD of the 3D 
building that can be generated with tools such as OSM2World...

My thesis is still in an initial phase but I'll have my method clear by 
February and I plan to submit in May-June.
The reason why I wrote this general mailing list and not to the 
osm-tagging list is that I'm very curious about any idea, critic on my 
thesis or GSoC project idea. So I hope this will open a discussion.

I was thinking to target the osm-tagging list but I'm not sure if that 
is the most appropriate. Could you advise me is this matter as well?

[1]: https://github.com/mablab/rpostgisLT/wiki/GSoC-2016
[2]: https://cran.r-project.org/package=rpostgisLT
[3]: http://www.sciencedirect.com/science/article/pii/S0198971516300436

Best regards,
-- 
Balázs Dukai
MSc Geomatics engineer candidate | MEng Landscape architect
The Hague, Netherlands
LinkedIn: www.linkedin.com/in/balazsdukai


More information about the dev mailing list