OpenStreetMap
My new eTrex arrived in the last week or so, and as the ordered rechargable batteries haven’t yet arrived I stuck a couple of Duracell batteries in and took that and my old GT-31 out with me for a ramble yesterday (or pub crawl - we were marking CAMRA’s Cider month by walking between 5 cider pubs over a 7.2 mile route). I wanted to compare the traces from the two.
So, this morning I loaded all the traces into JOSM, coloured the one set red, the other set blue and added the Bing imagery background. (Aside: at some point I’ll probably upload these traces, though I tend to only upload in date order and often only catch up at Christmas; I’ve captured a few images and put them on Flickr.)
Both GPSes were in the same pocket of my fleece jacket, except when I brought out the Garmin to show how OSM maps were displayed and included the POIs we were passing. Thanks to Talkytoaster for the easily added OSM map. I’ve yet to try other OSM maps, though I believe others are available.
Based on yesterday’s single test, in summary the GT-31 seems more accurate when I’m travelling on a bus or in a car (especially in the car). When walking across fields, the Garmin looks to generally be a little more accurate, though there was very little in it. The GT-31 creates a larger point cloud when in pubs (the Garmin was more likely to report a signal lost). The GT-31 was only set up to capture when moving above 1km/h (iirc) so the Garmin showed a dense point cloud when I was waiting for the bus home for 5 minutes that the GT-31 didn’t.
Discussion
Comment from Elkim on 21 October 2012 at 22:20
Side pocket for Etrex - bad idea. Better plase is on the top of backpack’s strap.
Comment from Dr Kludge on 22 October 2012 at 00:21
Your post brings up an issue that I have been thinking about. The arm chair mapper page makes it sound like GPS mapping can do no wrong. Your pictures of your traces and my experience point to revisions that need to be made to this article. There needs to be a similar article about issues with GPS mapping. Neither of these techniques are perfect but add to better maps.
I went out with two devices in order to GPS a new road that is being built in the Phoenix Arizona USA area.. I had already roughed in the road and multi use foot/cycle path and wanted to improve road. Moreover, the area is in a desert preserve. There were no buildings to provide echo problems or other GPS interference problems. Also both the units were on external battery sources. The Etrex Vista cx GPS unit was on the shoulder strap of my water backpack. I was taking pictures with the HTC phone GPS with keyboard mapper running. I averaged the resulting two traces while improving the roads in JOSM because there were wide variations in the two traces.
Things I now do, have experienced, or understand:
I zoom in as far as I can when tracing aerial imagery. If you are at zoom 0, there’s no way you can correctly place a point on Phoenix Arizona, for example. Note that the Yahoo aerials where not as hi res as Bing is. The difference in resolution could cause many accuracy problems while tracing imagery in the past compared with today’s available imagery. This could be one of the reasons that tracing has gotten such a bad name.
If you trace a GPS path at at zoom 0–some high level zoom–then you will have the same issues as tracing low res imagery.
Microsoft being Microsoft has its map failures along with its software failures. I don’t want to sound unthankful here. I want to thank MS for the use of their imagery as one of my mapping tools. If I trace, say, a building in the Phoenix area, I can get in nice and close. The building is nice and square. However, the low level image is not the same place on the earth as the newer 2011 images in the Phoenix area. I have to move the resulting trace around to where the building is in at higher zooms. This provides me with doubt about what I am doing at times. Not only are the there image registration issues, there are different years at super low zoom levels. The resulting map is still useful, however. I have had the chance to use OpenLayers to overlay to OSM tiles over other imagery. The traced OSM map still matches imagery from another source that is constant at all zoom levels and correctly registered to the world.
I tried tracing a new building in Phoenix by using a GPS trace. The GPS trace was useless because of building echos. I had to wait until better imagery came along before I could trace the building.
There are GPS failures that you need to be aware of just like you have to be aware of aerial imagery failures. The arm chair mapper page is a neglective fiction as it only point outs out the issues with aerial tracing.
The TIGER data import had a bunch of lower quality data. The TIGER data was good enough to get census worker’s “boots on the ground” to count people for a census but not good enough for maps without correcting the TIGER data. However, I cannot imagine trying to walk around the US with a GPS to start the US map. Fixing the TIGER import with imagery and supplemental GPS traces was a good solution for the US.
A mapper would have to buy professional grade GPS devices to have a better shot at accuracy. Even regular GPS mapping with consumer grade GPS devices is a great barrier to contributing to the map. However, in some cases that is all you have but it is good enough.
I now have doubts about what ground truth really is. ;-) However, I am able to do useful things with the tools that I have available to me. LOL. I don’t need ESRI or CAD tools. JOSM and Potlatch tools have been great. Commercial and expensive close sourced tools will not provide any better results!
Greg