WikiProject Belgium/Building and address import/SPW building and address import

From OpenStreetMap Wiki
Jump to navigation Jump to search

The GIM Tool is ready to import buildings with addresses in all of Belgium. This page will outline the specifics for the Wallonia datasets.

The import will follow the GRB import plan. The main differences need to be documented:

  • building types are provided in the source data, so they do not need to be derived as they are done in GRB
  • the addresses (from ICAR) and buildings (from PICC) are NOT linked together at the source. They were merged spatially. We should describe the process.

Known quality issues of PICC and ICAR

Issues mappers may notice

These are just the issues that are NOT addressed before the data is made available in the tool.

While the vast majority of information is correct, we know from experience and by comparison to on-the-ground observations that there are some errors in the PICC and ICAR databases. Always consider that some geometries on OSM might have been mapped by a local who knows the situation well. Don't replace existing geometries without making sure you're not destroying better surveyed work.

From what we know, PICC data is created/updated only from aerial imagery in some places, while other sources are used in some places. Since there are recent imagery available, you may have a nicer view of the building than the one that was used for digitizing the building in PICC, so always check the last imagery.

Some of the known errors are:

  • Incorrect shape: Sometimes the roof shape is mapped as if it was the shape on the ground (which a normal OSM building should be). Sometimes the shape could be considered correct if you consider it a "floating structure" (building:min_level=>1), but this is not available as attribute data in the PICC. There are also some errors with small buildings in backyards ( sheds, cabins, etc.) that are sometime hard to detect on imagery.
  • Not a building: In some rare cases, some structures in backyards such as covered wood piles are mistaken for buildings in PICC.
  • Razed building: A building that used to exist might still be in the PICC. Sometimes this can only be known by checking locally, sometimes this is already visible on the last available imagery.
  • New building: Of course, you can see new buildings on the ground (or even in the last imagery) that is still not present in PICC.
  • Incorrect ICAR / PICC address: We know of cases where the ICAR housenumbers were incorrect with respect to the on-the-ground situation.

Issues fixed by the tool

The source data has overlapping buildings. There seem to be two causes:

  • real adjacent buildings that aren't perfectly digitized
  • a single building that exists in more than one database version simultaneously

The second issue should be impossible: when a new version of a building is digitized, it should retain the same internal object ID and a newer datastamp. What happens is that the internal object ID is (accidentally?) changed, which makes it impossible for the database query to know that these objects are related. The logic applied to these buildings is programmed here.