We are aware of the Canvec interpolations and cleaning them up(residential areas with 100% coverage, we keep them in vast areas where new buildings might be built as to not miss any potential areas) so that process is on going...
On Thu, Jun 29, 2017 at 8:01 AM, Rory McCann <[email protected]> wrote: > Hi. > > I think there's been a misunderstand, maybe I'm not being clear. > > I did some quick analysis of the data & current OSM data. There are > about 622 trees which are inside a building. about 300 which are <1m > from the centre line of a road. > > For example there are some trees-inside-buildings around here ( > https://www.openstreetmap.org/#map=17/45.39017/-75.75801 ) and ( > https://www.openstreetmap.org/#map=16/45.4187/-75.7007 ). My JOSM > validator didn't flag "tree inside building", so you can't rely on that > to figure it out. > > There are "trees very close to road centreline" around here ( > https://www.openstreetmap.org/#map=18/45.28647/-75.70499 ) or ( > https://www.openstreetmap.org/#map=18/45.40001/-75.75015 ) > > For a dataset of 150k, these numbers are pretty good. > > When doing an import, it's important that someone does this sort of > work. That's part of the process of doing an import, to learn more about > the accuracy of the data you have, and to check how it would look after > you import it. You shouldn't just presume everything will fit together > 100%, someone needs to do this sort of analysis. > > I suggest you filter out the small number of suspicious trees, and > either manually enter them, change the OSM data, and/or contact the City > of Ottawa to tell them about possible errors in their data. > > BTW I noticed there are lots of address interpolation ways from a 2011 > CanVec import, and then lots of buildings with address tags from a > building import this year. The old address lines weren't removed, so I > think there are thousands (hundreds of thousands) of duplicate addresses > in Ottawa? Have you noticed this? This is a danger of doing an import > without looking at the existing OSM data. > > Doing some data analysis isn't a "mechanical edit", you're looking at > the data, not editing it. > > Rory > > On 28/06/17 17:13, James wrote: > >> Other than MANUALLY VERIFYING EACH AND EVERY TREE, there is no way to >> give a statistical analysis of the accuracy of the entire dataset. If we >> did it programatically we'd have to prove how the method of analysis is >> correct and would be probably be brushed off as being a "mechanical edit" >> thus invalid. If the need to verify each tree is not on a >> road/water/building this could be accomplished during the import(what I >> like to call a manual import: where the data is validated at the time of >> the importation of data as to the accuracy of the location and not just a >> massive dump of data in one pass) >> >> On Wed, Jun 28, 2017 at 11:02 AM, Rory McCann <[email protected] >> <mailto:[email protected]>> wrote: >> >> On 28/06/17 16:53, Kyle Nuttall wrote: >> >> I am still a little confused by what you mean. Obviously if >> there's a >> tree in the middle of a building, that means something is not >> right. >> But what I was trying to say was that the trees in this dataset >> will >> only appear where an actual tree is in real life. There shouldn't >> be >> any cases of rogue trees in the middle of rivers or buildings >> because >> a tree just wouldn't be there in real life. >> >> >> What if OSM data is wrong? What if OSM says "The river bank goes up to >> here" and it doesn't, and that would make the tree in the river. >> Sometimes people just upload data into OSM without checking for things >> like this, and you get these weird data problems. You should try to >> figure where (if at all) this happens, and see if the problem is OSM >> or >> the official data. >> >> You're also presuming the official data is 100% guaranteed to have the >> correct location. If they are 99.9% accurate, then there are 150 wrong >> locations. 99.99% accurate, 15 wrong locations. >> >> If you do this sort of analysis you can find out just how accurate the >> official data is, and also prevent making the map be full of these >> mistakes. >> >> >> >> _______________________________________________ >> Imports mailing list >> [email protected] >> <mailto:[email protected]> >> https://lists.openstreetmap.org/listinfo/imports >> <https://lists.openstreetmap.org/listinfo/imports> >> >> >> >> >> -- >> 外に遊びに行こう！ >> >> >> _______________________________________________ >> Imports mailing list >> [email protected] >> https://lists.openstreetmap.org/listinfo/imports >> >> > > > _______________________________________________ > Imports mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/imports > -- 外に遊びに行こう！
_______________________________________________ Talk-ca mailing list [email protected] https://lists.openstreetmap.org/listinfo/talk-ca
