Hi Christoph, For data accuracy:
Yes, our goal is to automate the tagging for flat-roofed and unobstructed buildings. Sloped or uneven roofs and trees can be identified in the JOSM hillshade tileset [0]. Mappers will then refer to the raw LIDAR or street level imagery to determine the correct height. We want to prepare for mappers a complete set of tools and imagery to create an accurate dataset. We believe the result will be up to the standards of height data elsewhere in OSM. It should also be better than a mapper on foot could estimate without special equipment. For the community: I don't believe that this import will discourage community mapping. I used the Overpass API and found only 424 edit actions since 2012 adding a height tag to a building in San Francisco. Details are on GitHub here [1]. Two users are responsible for 65% of these building height edits. There are only 10 users total who have added a building height tag to more than 2 buildings. I have messaged all these users, inviting comments on the wiki. These results suggest to me that heights are not a current focus of mapping in the affected area. I do think that getting started with building heights will encourage further mapping. A complete 3D dataset like in New York City is a great showcase for OSM data and attracts new users. Getting to a similar level of detail in San Francisco is a compelling goal for the local community. We've had offers to help from local meetup groups and companies that are excited about the project. Brandon [0] https://wiki.openstreetmap.org/wiki/San_Francisco_Building_Height_Import#Failure_Modes [1] https://github.com/osmlab/sf_building_height_import/issues/21 On Thu, Nov 17, 2016 at 9:41 AM, Christoph Hormann <chris_horm...@gmx.de> wrote: > > First of all thanks for doing a more elaborate preparation than back in > May. The whole process is now much clearer. > > To verify my understanding: the height values you assign are the median > height values of the city footprint data set when there is a matching > footprint within the area percentage cutoff chosen. > > This median height value is the median height in the 0.5m gridded data > set within the city dataset footprint that has been calculated as the > difference between a gridded first reflector data set and a gridded > ground level data set, both derived from the raw LIDAR data. > > I am sure this often leads to fairly reasonable results, in particular > with flat top buildings and flat ground with no significant structures > except the buildings but it does not appear to be a really good > approach in principle. > > Sources of error here are not only the systematic error with non-flat > roofs, the footprint mismatch and obstructions of the roof. Likewise > important are the inaccuracies introduced by the grid sampling step and > ambiguities in the ground level definition like plants, cars and > non-building structures. > > These technical things aside i am not sure it is a good idea to enter > this kind of data in OSM. This likely won't encourage community > mapping and it will be of little gain for data users. Producing a > separate point data set with the height values that can be easily > matched with the OSM geometries would IMO be better. You could then > also replace the fixed geometry matching cutoff with a reliability > attribute and data users could decide how strict they want to be in > that regard. > > -- > Christoph Hormann > http://www.imagico.de/ > > _______________________________________________ > Imports mailing list > Imports@openstreetmap.org > https://lists.openstreetmap.org/listinfo/imports _______________________________________________ Imports mailing list Imports@openstreetmap.org https://lists.openstreetmap.org/listinfo/imports
