Here are Pasadena's buildings

[jschleuss]

Pasadena's buildings were the only ones in which most buildings were broken into several parts. To overcome this I've dissolved all features by the AIN (Assessor Identification Number) and then converted Multipart to Singleparts via QGIS.

Initial feature count: 113,309 After dissolve: 55,840

All buildings with a source of "Pasadena" vs. LARIAC2 or LARIAC4 contain no height or elevation data.

Here are the pasadena buildings: http://latimes-graphics-media.s3.amazonaws.com/jon-temp/pasadena.zip

In theory we should be able to just change the URL on this line once we're ready to go.

@maning can you check this file? It should be good.

Also, I find it terrifying that Pasadena's buildings look like an angler fish...

[maning]

I converted the Pasadena data into OSM files and reviewed at least 20 files. Data is here: https://github.com/maning/labuildings/blob/pasadena_test/osm_pasadena.zip

Notes below:

Alignment to imagery is very good

44/17535 buildings have errors flagged by JOSM validator

Specific errors are

• Duplicated nodes - not exactly overlapping but nodes that are very close probably caused by the merging.

• Duplicated ways

• Building inside building - in the example below, the interior way has a building tag which was also part of a multipolygon relation

• Self intersections

• Crossing buildings

• Ways with same position

The above errors were automatically flagged by JOSM validator and can be fixed right away.

Overlapping buildings from existing OSM data.

Out of the 55k buildings to be imported, 9,351 overlapped with existing OSM data mostly south of Pasadena (OSM-Magenta, Pasadena import-Cyan, Overlaps-Red dots).

Next actions

• Most common errors can be easily fixed by experienced importer/user so I'm not very concerned about this but if we can fix this programatically before import is better.
• What is more difficult is the overlaps from existing buildings in OSM, this can be tedious to do manually. Let's think of ideas on a conflation workflow.

@jschleuss @almccon @cityhubla

[almccon]

Since most of Pasadena south of the 210 freeway is already mapped maybe we just draw a mask over that area and don't try to import it?

[jschleuss]

@maning @almccon

I'd like to continue with the import south of the freeway because our data will in some cases be better than what you can see in imagery, since it's pulled from stereo imagery. It should, in theory, have the building contours that would traditionally be covered up by trees in satellite imagery.

We also have all the building data associated with each property, such as building type, units, year built, AIN, building ID, height and elevation. I think those attributes make the data extremely valuable.

It looks like one user has traced most of those buildings. They're really well done, so I've reached out asking them for input into this discussion.

I'm thinking we replace geometry whenever the import data "appears" to be better than what's already there. If what's there is better than the import we could copy/paste the tags over so the building attributes are improved.

[jschleuss]

Oh, also meant to say that it would be nice to fix those errors before we do the import, though they could be fixed manually by advanced users.

[almccon]

Oh, and sorry if this has already been discussed, but are the multipart Pasadena buildings something we could transform into the OSM 3D building schema? Because that would be even cooler than just adding building heights!

[geochetan]

According to OpenStreetMap Simple 3D building tagging scheme, to consider 3D rendering all building:part=yes polygons should lie inside building=* outline.

green = building:part and red = building=*

Observations

• The shape file doesn't have height or building:levels information. These two features are necessary to decide which building part posses more height in a building. From taginfo, approximately 82% of building:part have tagged with either height or building:levels in OpenStreetMap.
• Taking an example of a building to make it in 3D from the shape file.

This building has 3 small parts. These building parts share common edge with building outline. Here we have to do two things:

1. Add building:part=yes tag to 3 parts.
2. Include all 3 parts inside the building outline. To do that we need to split the outline and trace on top building parts like below.

• There is no attribute in the shape file to covert it as building:part.

Next actions

• The ideal thing is to merge all single building parts as one building and follow the same process of import.

@jschleuss @almccon @maning

[almccon]

Thanks, @chtnha! I forgot that the Pasadena data was missing heights and levels. In that case, yeah, there's no 3D information worth extracting, so we should stick with the original plan of merging all the shapes.