<div dir="ltr"><div class="gmail_extra"><br><div class="gmail_quote">On 7 February 2017 at 23:52, Andrzej Popowski <span dir="ltr"><<a href="mailto:popej@poczta.onet.pl" target="_blank">popej@poczta.onet.pl</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div id=":1rq" class="a3s aXjCH m15a1ac72c23f7df9">I like your ideas about is_in and urban areas. But I have doubts about practical results. IMO areas like landuse=residential or place=city/town aren't mapped well in OSM data. So even if you add some new features to mkgmap, it won't be useful, because input data is not precise.<br>
<br>
Some consideration about urban area. I think using buildings or house numbers as indication of urban area won't be reliable. Mostly because these are data from imports, so some cities get them while others still do without. Probably highway's junctions (crossroads) could be the best indication of urban area. Roads are usually mapped correctly and number of junctions shows complexity of road network.</div></blockquote></div><br></div><div class="gmail_extra">You have a good point there - I also think residential and city/town are not that well mapped. Housenumbers are pretty well mapped in many european countries though - but not in the rest of the world. Buildings I would guess are sufficiently mapped in most european countries in OSM. <br></div><div class="gmail_extra">Road junctions have one problem - actually residential areas (as in where people are living) have most junctions, while city centers / commercial areas feature much bigger buildings and less intersections. Maybe the kind of roads in such areas would help to better identify them. Urban areas would be easy to find however - take the amount of crossings from highway=track/path/cycleway vs highway=residential/primary-secondary-tertiary-unclassified and the picture should be very clear.<br><br></div><div class="gmail_extra">I think both methods will allow for an improvement in routing as well as an improvement to exclude clutter from buildings in maps if we can derive a classification level of urbanisation from it and feed it to mkgmap like bounds and sea files.<br></div><div class="gmail_extra">My idea would be there should be an analysis of 400x400m patches - then join them and find out how big the area becomes - any areas that have more than 3-4 connected patches can then be classified as urban in the "urban_bounds" file - if it's one or two of such patches - I'ld rather keep the buildings in my maps and also still consider it rural (because for a tiny mountain village - I'ld like to keep them - however for a mountain town like Zermatt, Ischgl or their like - I'ld prefer to remove them - while for big cities is essential to remove the buildings.<br></div><div class="gmail_extra"><br clear="all"><br>-- <br><div class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div><div>Felix Hartman - Openmtbmap.org & VeloMap.org<br></div>Schusterbergweg 32/8<br></div><div>6020 Innsbruck<br></div></div>Austria - Österreich</div></div></div></div>
</div></div>