Currently I'm using a Samsung GSII to make GPX files of unmapped streets and alleyways in Hanoi. My problem is that in Hanoi there are a lot of narrow alleyways surrounded by 4-6 story buildings, and the GPS can't get closer than 50m accuracy in these conditions. I end up doing more guesswork than I'd like when I'm mapping the alleyways. Are there any Android apps that combine pedometers and a compass to make approximate traces of walking without using the GPS? It needn't be perfect - it would just allow me to get a combination of aerial photography, GPS and this to get a better guess of the alleyways. asked 10 Apr '12, 14:59 samgrahk |
I do remember someone posting "maps" of a US university that they'd done this way on either one of the OSM lists or IRC channels - the results were better than I'd have expected. A quick web search doesn't find exactly that example, but does turn up some possible answers to your question. answered 12 Apr '12, 11:23 SomeoneElse ♦ Thanks for this. It looks like this is a common problem for car GPS navigation, and with a fairly brief look, it seems there aren't any apps. These seem to be looking at using the accelerometer to sense turns in the car. I'll keep looking, but this is a nice lead...
(13 Apr '12, 07:06)
samgrahk
|
With some GPS devices (Vista HCX for example) you can get it to take a number of readings and average a point, so maybe if the alleys are straight using a point got from open areas at the ends could work, or a compass bearing from one. One way to get an average with an android is to stay still for a while and the resulting blob in the trace is an average.
That'll help, but with tall buildings and (I'm assuming) no EGNOS equivalent in Hanoi, an "averaged" point may still be displaced to one side. The "if you can, draw a straight line between points in open areas" idea that you mention is probably the best idea available.
Taking an accurate measurement at both ends works for some alleys, but you can get an idea of the complexity of some areas by looking at this section: http://www.openstreetmap.org/?lat=21.03863&lon=105.8198&zoom=17&layers=M . (I didn't map these, but I've checked a few of them out for accuracy, and they seem to be largely right.)