The major issue with the last set of .osm files I generated using the road-matcher matches was that it included a significant amount of streets that were already in OSM but road matcher failed to match. Importing those .osm files uneditted would give a fair amount of duplicate ways (that would need to be manually cleaned up).
I tried a slightly different approach this time. Road matcher categories roads from the two datasets into three main categories. Matched (It can determine the road is in both sets), Standalone (It can determine the road is only in one set) and unknown (all the roads that don't fit into the other two sets). The last .osm file used the Roadmatcher matches to exclude geobase roads from the import. This one only includes the Geobase standalone roads. We could: -Generate the Geobase roads that are standalone compared with the existing OSM data. -Import these roads into OSM. We would then have a mostly complete road network with very few duplicates without having to delete any OSM data. -We could then manually look for 'missed' roads in JOSM by overlying the complete Geobase roadnetwork and manually add these in. This approach would not add street names to any of the unamed OSM roads. We might want to consider deleting the unamed streets and replacing them with the geobase versions. This approach would also create two classes of OSM streets, those with geobase nids and those without, but we might end up in that situation anyway. I think it would be possible to apply the Geobase street addresses to exisiting OSM streets provided that the streets are named even if those streets lack an nid. We could write a script that finds the OSM way with that name, validates that it is at a lat/lon close to where goebase thinks it is, then add in a node for the address point specified by geobase. The sample .osm file for the 030M05, Hamilton area can be found at http://www.mediafire.com/?ye012dpmpbc I am open to any ideas anyone has on how imrove RoadMatcher's success rate at turning unknows to accurate matches. Steve _______________________________________________ Talk-ca mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk-ca
