I run a website that collects GPS trail (track) information of trails
and in turn distributes and displays this data to users. Now, as many
of you know, GPSs are not necessarily 100% accurate - take a couple
peoples tracks that mapped the same trail then overlay them, and,
while they are usually pretty close, it often looks like spaghetti.
I want to take all this polyline data, jam it into some sort of
software, and have it return a nice "averaged trail networks" data
layer.
Now my software of choice to use with my GPS is Topofusion and it does
exactly what I want, but it does not seem to handle lots of data very
well / is pretty slow / crashes some.
Take a look: http://www.topofusion.com/network.php
Basically, I need an enterprise version of that.
Any help or ideas would be much appreciated!
Thanks,
Nate
Here is an explanation on data smoothing techniques.
http://documents.wolfram.com/applications/eda/SmoothingDataFillingMissingDataAndNonparametricFitting.html
Probably, the most general purpose curve fitting algorithm is
Levenberg-Marquardt.
There are also genetic algorithms for this sort of thing.
I guess you will be unhappy with a simple moving average unless you
Hahn the tails, and you are likely to have general problems with
arbitrary data because it won't create invertible functions (e.g. a
simple loop).