|
Hi, i recently did a few gps traces of my local area and town and when i came to edit the traces on potlatch i found that a few of the roads were slightly off, maybe a cm to the left of the trace e.t.c. I'm wondering whether i should shift the road so it matches my trace or leave it as it is and assume the trace might be off slightly...? Also, the aerial imagery only shows up at something like 5 zooms out and makes things quite hard to make out. I've been using Motion X GPS on the iphone to create new tracks whilst ive been driving around if thats any help... Cheers. |
|
It is very hard to judge the relative reliability of aerial imagery and GPS tracks in general as both can be great or next to useless. There are however a few things you can try: See if your GPS tracks match the tracks by other mappers. If several GPS tracks taken over a few days are consistent with each outer then they are most likely correct and the aerial imagery is warped / shifted. If there are not other tracks yet you can revisit the street the next day and take a second track. Again if it matches the first track it is most likely more accurate then the aerial imagery. You can also check if different sources of aerial imagery are in agreement. If they are not at least one of the layers in wrong. If you don't fell certain enough to make the call after these test you should talk to other mappers in you area to find out what they use as authoritative source. 1
You should also try another editor which allows to zoom in without loosing the background even if there is no good imagery available.
(09 Mar '11, 23:06)
dieterdreist
|
|
Neither! Assuming you use normal consumer GPS its error value is something between 5-50 meters. In cities and in mountains where there are large vertical obstacles it can be much more. The aerial imagery can be precise if the stitching of the tiles is done well. I've seen places where a road is cut in half and two parts of it goes in two different directions... Also tiles can be OK, but the whole set can be shifted in some direction, or distorted. For best results if you want reliable data you should combine both methods. Gather some amount of GPS tracks, then average them to average the error. Put them over the aerial imagery. Search the tracks layer for places that you can identify - like crossroads - you can see traces crossing each other at right angle. See if that point matches at the aerial imagery layer. Repeat the process for several places and you can easily see if the aerial imagery is shifted. In general when working on a city I prefer aerial imagery (Bing). The GPS tracks web is too dense there and in bigger cities you can distinguish only the main roads and everything else is a mess. When working on mountain I prefer to combine GPS with aerial (if available). Always try to use at least several GPS traces on the same road, because it is hard to find an exact trace (and you don't know which one of these you have is correct). When working on plains, GPS should be enough (if reception is fine). Sadly reception information is not stored in the track file. 1
You are right that both sources can have errors. BUT: 50 metres GPS error if you have a good fix is not probable. The usual error will be between 3 and 15 metres (depending on the surrounding, device+antenna and weather). Aerial Imagery on the other hand, if produced professionally by a government agency, will usually be OK for position and form (e.g. in Italy we got the permission to use imagery from the agency for the environment and its positional accuracy is much better compared to the commercial products like the ones used in Bing and GMap, depite the lower resolution).
(11 Mar '11, 17:46)
dieterdreist
2
50 meters is more than possible when you are in city. The device says "good fix, 15 satellites, blabla", but in fact most of the signals you got are reflected from surrounding buildings. One good example is that if you enter a building - there all you receive is reflected and you can see each second how your position is moved 200-300 m around.
(13 Mar '11, 07:30)
ivanatora
maybe you should change your device (or antenna)
(13 Mar '11, 11:55)
dieterdreist
|
|
Besides the offset, a less common but more disturbing error with the images is that sometimes they present dead straight roads (or other features) having curves in the middle. There's a technical reason why that happens sometimes - they're assembled from a multitude of pictures taken from different locations and the earth isn't flat. In those cases you should draw the shape as you saw it on the ground. Bear in mind though, that, as others already implied, the gps log, too, can have bends in places where you travelled in a straight line. 1
When I'm going out to verify what I have traced I always bring a camera with me and if I see something like this I take a picture to remind me later. In my area the big problem is the imagery is out of date due to recent developments and sometimes roads have actually been straightened that were once curved. If you take your GPS with you as well you can use JOSM to geotag each photograph so you know exactly where you took it. Quite useful.
(11 Feb '11, 12:22)
Borbus
3
Basically I think the key point is don't solely rely on one source of data. The main sources of data for me are my own GPS traces, other's GPS traces, Bing imagery and my eye/camera when on the ground. Use them all.
(11 Feb '11, 12:23)
Borbus
|
|
Depends on the GPS trace, and depends on the imagery. Australia's Nearmap imagery is renowned for being particularly accurate: generally less than 1m. Your GPS might, on a good day be as accurate as 3m. On a bad day it might be 20m or even worse. Your GPS being part of a mobile phone, and inside a car at that, I wouldn't expect too much from it. |
|
get a few traces of same the area,either yours or others, you should get a good average then align the background image to agree (hold the space bar then drag in potlach2) then you can map the average of the traces and the image,just check it when you have scrolled a bit. 2
yes, and always check for uploaded traces from other users (load gps-data). The sum of all available traces usually gives a good impression where roads actually run. Don't hesitate to upload your own traces (if you had sufficient reception), even if there are already other traces online available for the roads.
(13 Mar '11, 13:05)
dieterdreist
I now upload traces even in well mapped areas as each trace improves the average.
(13 Mar '11, 17:52)
andy mackey
|
|
More art than science. If there are a lot of buildings near, trust the aerial image. If in the open, trust the gps. A common problem is where the GPS traces came from. Damn few OSM contributors will be driving down the center lane in a BMW. Here, most (me) are on a bicycle in the far left lane, and do the same every trace. This can throw the trace off 30 m. |
|
The wiki page on Yahoo! Aerial Imagery/Accuracy, should answer your question. "If you think you've discovered an area where yahoo imagery is offset from reality, please bear in mind first of all that it's actually far more likely that your GPS trace is wrong!" ...read more This page also has a section where we can list examples of Yahoo! imagery inaccuracy. So far not many. Please edit the page to add examples. This might help us get a more concrete feel the extent of inaccuracy. 1
Don't know for which parts of the world this is (was) true, surely not for the Rome-area. All Yahoo!-imagery has (had?) a systematical offset of more then 20 metres.
(11 Mar '11, 17:48)
dieterdreist
|
You are talking about Yahoo! imagery I presume? You should mention this in the question. You might also mention where your local area is (which country)
Also it depends on the date of aerial layers. If these are too elder, so GPS tracks. But GPS has an error of 10 meters. So I think the best is to compare and evaluate yoursafe. Simple answers don't exist in maps ;-)
There is a page on accuracy problems in the Wiki: http://wiki.openstreetmap.org/wiki/Accuracy That may partially answer your question.
Another interesting page might be http://wiki.openstreetmap.org/wiki/True_Offset_Process