OpenStreetMap
I have been comparing maps and satellite images from both Google, Yahoo, and Bing and for the same location I found different lat/lon values. I saw a reference to the Bing accuracy in wiki. Which of the six combination is most accurate or does it depend on the location?
I also found a reference to the accuracy of GPS traces. I understood that it very much depends on your receiver, the day (if not taken on the day they can be quite off). What is the best way to get accurate traces? Is there a way to create a "average" trace based on a number of trace? What should I rely on more, maps / images or GPS traces?
Дийцар
Коммент maxolasersquad 1 August 2011 01:23
The accuracy of the bing imagery where I live (South-East USA) is quite good. I tested this by aligning roads with the Bing imagery. I then used OsmAnd on my Android-powered phone to see how close me standing on these roads looked on the map. After checking various times I determined that the Bing imagery was close enough that one would need a very expensive GPS device to find any error. As far as it goes for one using the maps to get around in any reasonable scenario, it is good enough.
Коммент Chaos99 1 August 2011 06:14
There are different kinds of errors wich all have to be taken into account. Image maps tend to be shifted in some direction by up to 50m for absolute position, but are very good in terms of relation between points. So if you want to get the shape of a park or the size of a pool, go for the image maps (that is: bing, you know you are not allowed to use google, right?).
GPS traces are also error prone, but at least you normally know how much. You can't take the "+/-2m" literally, you might still be off by 10m, but it is a start. As every single point in a gps track might have a different error, those traces are not good for accurate shapes or relations, but within a margin they give you a good absolute position. You'll get even more accurate data when averaging over multiple tracks. Best done with different devices, on different days.
There are of course errors even averaging will not correct. A big building which blocks half the sky and reflects the other half will always interfere with the gps signal. A straight road in front of such a building will have a slight bend, even after averaging. Use common sense and local knowledge (or image maps) to know which roads are straight and which are not.
Hope that helps a bit ...
Коммент z-dude 1 August 2011 07:42
You'll find that outside city limits, the tiling seems to be done like a jigsaw puzzle, one tile shifted to match the next one. In the middle of a jungle such as at the Guyana border, the imagery can be off by 100 meters.
Коммент Anna_AG 1 August 2011 21:10
Yep - all true - BING can be offset, ( I have pre built offsets for Kaduna ( northern Nigeria in Josm, backed up / confirmed by dozens of GPS traces taken over many days ).
Yes GPS can be off, and Garmins sometimes decide to offset themselves for a few minutes by several meters and then re correct themselves.
Really the trick is to have many GPS traces of the same primary highways, uploaded by different users, aggregated over periods of perhaps several years. In this way a fairly good average can be built up, and confirmed by Bing. At least with Bing, the offset will be reasonably constant over a certain area, so once you have an aggregate of GPS traces you can offset the Bing by a reasonable degree.
We should note that if you are using a GPS map derived from OSM, then GPS the user is navigating with will just as likely have a GPS offset, even if the map is 100% accurate, so I guess the word is always caution!
The map is probably about right, how good is ones receiver GPS ?This is possibly why many GPS units also offer the option of locking to map to 'correct' possible location errors.
Коммент PsJp 1 August 2011 21:46
Thanks for all the comments. bri g mentioned that it is possible to offset the map in JOSM. Is something similar possible in Potlatch 2?
Does local authorities maintain points with an accurate lat/lon and if so does anyone have experience with using such a point to adjust the offset?
Then again, the French were off significantly when determining the equator, but that was a more than a hundred years ago :-)
Коммент z-dude 3 August 2011 05:51
yes, there will be survey benchmarks in a city for example.
http://en.wikipedia.org/wiki/Benchmark_(surveying)
Also, when I grew up, property was staked out, and there'd be a peg in the ground specifying a property line corner.