On Wed, Dec 22, 2010 at 6:27 AM, M∡rtin Koppenhoefer <dieterdre...@gmail.com> wrote: > yes, in you example you would have 100 wrong streets. I'm not > believing your numbers btw.: I doubt that you can only visit and map > 10 streets with the effort you have to put 1000 streets from > orthofotos (1%). Even if this ratio was only 10% (in my experience > mapping takes as long as surveying, which would result in 50% for no > survey at all) I would prefer 100 reliable streets to a thousand of > which a hundred are wrong. If there is no information, this is at > least reliable in the sense that you know you can't rely on it ;-)
Well, you're welcome to your preference. As to my numbers being wrong, it's probably the other way. When I'm on a roll, I'm probably mapping a street from imagery every 5 seconds or so. Click, click, click, "r", done. I don't know how long it would take to drive up and down it (surveyors recommend at least two GPS passes don't they?), then import, then trace, then tag. A lot more than 50 seconds, anyway. And it's not as if GPS traces give brilliant results anyway. > A very good map can't be done just from orthophotos. You thought I was saying the opposite? I would say a perfectly usable map can be made just from tracing imagery: you'll have the roads, with basic category distinctions, plus some footpaths, bike paths, parks, carparks, buildings etc etc. There will be occasional mistakes like driveways mapped as roads, incorrect junctions (ie, two roads that pass near each other but for some reason don't join), missing gates etc. And of course you won't have names. For the way I use maps, that's actually pretty acceptable: I tend to load them on a GPS and use them for navigation to a known point. So, knowing about roads that connect places is far more important than knowing what they're called or speed limits or whatever. (Also, since I cycle, a lot of that extra information about roads is irrelevant). Of course, we don't build maps just for our own individual preferences, but we're certainly biased towards including the information that interests us personally. Steve _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk