Re: [Talk-us] Import accurate streets in Portland, Oregon - Trimet
Questions for the talk-us: 1) How do you feel about some kind of automatic editing or (selective) import? Other methods? Hi PJ, The actual method of import doesn't matter - only the quality of the result is important. Roads and trails should properly connect to existing data with no duplicates. 2) How do I go about doing that? (I heard of something called RoadMatcher)? I just did an import for Murray County, OK last week. I did a remove / replace operation because most roads had never been edited, and many road names had been changed. Something like RoadMatcher would be much better because Existing OSM attributes, history, and edits would be preserved without a manual copy operation. Easy to copy correct geometry onto existing road. I glanced at RoadMatcher, but I am not familiar with the toolset that goes with it. Expect some manual work with any import to properly stitch new or updated roads onto any existing roads. There is some interest in writing an OSM tool for JOSM or another editor that will streamline all the operations, but the timeline is not known. 3) Any other input? Some types of data fit into OSM better than others. Natural imports are roads, trails, bike routes, park and golf course boundaries (which are often difficult or impossible to survey by hand, and existing park boundaries are easy to spot on the map or detect in the OSM dataset). Having updated roads, trails, and routes in OSM allow other navigation tools to work with the data and provide seamless routing to the rest of the country. Public transit routes also fit in OSM, but since Trimet will be using a multi-modal approach, it is best to postpone bringing this data in last, if at all, unless it can be automated so that it is always up to date. ___ Talk-us mailing list Talk-us@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-us
Re: [Talk-us] Import accurate streets in Portland, Oregon - Trimet
On Wed, 2011-01-12 at 22:37 -0800, PJ Houser wrote: In Portland, Oregon, the region's transit agency Trimet (http://www.trimet.org) is transitioning to an open-source trip planner (http://maps5.trimet.org/otp/). The trip planner will use OSM data for a multi-modal approach - walking, biking, driving - and Trimet's route data. However, Portland data is not accurate enough nor complete enough yet in OpenStreetMap, so my job is to help fix that with Trimet and the community. The roads are often incorrect or do not intersect properly, making consistently good routing impossible. Bike routes and off-road trails are seriously lacking. I think there are a couple of distinct problems here. With the things that are simply lacking from OSM, I think a careful manual import makes sense. It would be nice to get a sense of the scale of the problem, like how many trails we're talking about in the metro area. If we're talking about 100, a manual import makes sense, but 100,000, probably not. Similar to that problem is getting new streets that weren't imported with TIGER2005 in to the data set. Coming up with that set can probably be done in a fairly automated fashion. But, perhaps doing it visually would be good enough: http://greenvilleopenmap.info/StreetCompare.jpg Another concern is what ongoing efforts there will be each time new data gets released. We should at least keep in mind what will have to be done next year at this time. Especially for a trip planner, I'm not sure that that accuracy is really all that important. Here's a route from where Google says my old apartment was to where it actually is (~420 feet off according to them): http://goo.gl/LjxWv That's basically a precision problem in their geocoder, and is the kind of problem that will pop up if OSM streets don't closely match what's coming out of your geocoder. It's something to work on, but it doesn't seem like an urgent problem to fix. I think people are used to the endpoints of their route being a bit fuzzy. No matter what gets done, we'll need something to convert over to the OSM formats. That's probably the most important thing to be looking at, first. OpenTripPlanner looks really cool, btw! -- Dave ___ Talk-us mailing list Talk-us@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-us
Re: [Talk-us] Import accurate streets in Portland, Oregon - Trimet
I did a quick ogr2osm conversion on the street data from here: ftp://ftp.oregonmetro.gov/odi/Streets.zip I stuck it up here: http://daveh.dev.openstreetmap.org/portland/ I loaded it in (JOSM takes ~1.2GB of RAM doing this), along with some downloaded OSM data. Then I set the inactive color to a nice, bright pink, and selected the downloaded OSM data layer. The streets that aren't in OSM already stand out pretty clearly: http://daveh.dev.openstreetmap.org/portland/josm-ping.png Wherever you see pink, there's potentially a problem. It might be a misaligned or missing OSM road. This doesn't guarantee the names are right or anything, but it's at least nice to get a handle on how complete the OSM data is and how well it coincides with Metro's. I'm fairly impressed with how OSM stacks up. It's also neat how much of the missing stuff has actually been traced from imagery, but just needs names stuck on it. If this looks like something that might be usable for doing a manual import of new streets, we can probably work on a nicer JOSM style file to make the process even easier visually. -- Dave ___ Talk-us mailing list Talk-us@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-us