Re: [Gpsbabel-misc] extreme simplification

[Ron P. <ro...@pa...>]

Vidar Gundersen wrote:
>   -x simplify,count=1   Last location
>   
You might find that this is not consistently the case.  There's nothing 
in the code that says it can't be the first location.  Which one it is 
is essentially random.

> What is the middle point?
> Is it where half the distance was travelled,
> or the halftime position?
>   
Neither.  It's not even easy to define what it is, in fact.  In a sense, 
it's the point, not counting the endpoints, that made the single biggest 
contribution to the shape of the route.  In reality, due to various 
assumptions inherent in the algorithm I use to simplify routes, it's not 
necessarily the very best point that could have been picked to represent 
the route, but it's probably somewhere close.

As an example, say I have a route from Street Atlas that takes me from 
my home in Fort Wayne, Indiana to Toledo, Ohio.  That route would follow 
Interstate 69 roughly north about 40 miles until it gets to Interstate 
80, then follow Interstate 80 east about 80 miles to Toledo.  The route, 
then, is roughly L-shaped, with 1/3 of the trip going north and the 
other 2/3 of the trip going east.  It would consist of several hundred 
points, originally, but after simplification to three points, you'd 
probably find that the "middle" point was somewhere near the corner of 
the L.  That's not halfway by distance or by time, but it's the most 
important point in the route.  (In fact, I just checked and that's 
exactly what happens.)

> Back to my above tracklog inventory: I'd like to
> create a GPX file with one waypoint for each tracklog,
> where the name of each waypoint is the file name of the
> original tracklog (where the position should ideally be
> a geometric average, but the endpoint is also useful).
>   
I think your best bet is to try to compute the centroid yourself.  The 
simplify filter won't be of much use in that endeavor, except perhaps to 
winnow the route down to a number of points that's more amenable to 
processing with Perl or Python.