[Patch] Gpx tracks: speed outliers and coloring

[daganzdaanda]

I like to colour my gpx tracks according to speed, using the dynamic colour range.
Sometimes, my gpx tracks have errors in the recorded speed (e.g. when leaving a building, there might be a "jump" in the data).
In that case, the dynamic range is distorted and gets less informative.

Would it be possible to cut off the outliers for the dynamic coloring range?
Ignore the highest and lowest x% of speed values, painting them either grey or pink/red.
What would the best value for x be? 2?

[bastiK]

Could you attach one or more sample tracks?

[daganzdaanda]

Here is one with a weird hiccup: ​https://gist.github.com/daganzdaanda/af2e861e3c1cc31bed29/raw/booths_rock.gpx

[simon04]

attachment:11338.patch​ takes the 5% and 95% percentile as lower/upper speed value.

Due to the required sorting, the complexity changes from O(n) to O(n log(n)) where n = #waypoints. What do you think?

[simon04]

Postpone. Feedback is welcome :).

[wiktorn]

The only other approach that's possible instead of using percentiles is to compute standard deviation and average, and use max/min capped by (average +/- 3 stddev).

This doesn't require sorting and assuming normal-like distribution, this should cover 99.8%, though I haven't tested GPX speeds against normal distribution. But I think that it might suffice.

[simon04]

In 9234/josm:

fix #11338 - Gpx tracks: speed outliers and coloring