[OSM-talk] GPX trackpoint stacks

[Nick Hill]

I have been creating queries to determine statistics of redundancy in 
the GPX trackpoints.

At one point, we have a stack of 11385 gpx points in exactly the same place.

There are 1259115 in stacks of 6 or more,
2533037 in stacks of 3 or more
6497729 in stacks of more than one

Quantising the world into 0.000003 degree squares (about 30cm, 1 
imperial foot lat), and counting those trackpoints where there are more 
than one in a square, yields a redundancy of:
7297894 out of a total number of tiles 2655121 with more than one point.

Given that 6.5m are in stacks of more than one, and 7.3m points share 
the same tile, then either the same track points have been uploaded 
repeatedly, OR GPS receivers have some form of rounding function causing 
GPX points to snap to grid. It could be the case that GPS receivers 
generally use single precision for their internal representation.

Those additional trackpoints could be useful for advanced data analysis 
such as directionality of travel and road class but are not useful for 
drawing in streets. Sampling to 30cm gives better than +/-21cm accuracy.

...all out of a total of 29.7m trackpoints.