I started to play with location data and remembered my google location data which are collected since 2011 by my mobile.
I was interested in plotting all my moves into a map, in a best scenario as a heatmap.
That worked out fine with manually created (e.g. gps device) data but very bad with google location data. I poked around to find if it is possible to fix the errors (semi)automated.
As far as i tested now, it is not (easy) possible:
To eliminate the lazy (batterysafe) location aquisition by 'known' gsm and wifi locations (with often wrong/old/moving positions) which create this anoying rays into nowhere (scroll!) i needed to find and remove them.
It is interesting to know that i do know in deed some of the 'nowhere' locations. :-) One (shown in the 1st map pictured below) is the location where i used a mobile wifi2gsm router at the first time. It looks like google acquires the position of a new, never noted wifi through a connected device with position (my mobile with gps chip) and connects this position to the ESSID/Mac of the wifi network in a database and does not take into account that wifis may move.
I had some similar cases years ago, one in train with wifi and one as we moved into a new flat - we have been able to hack a ingress portal located just in front of the old flat, if we have been connected to the (same) wifi hardware in the new flat.
I consider that this is a known problem to google, but they dont care.
I started to compute the distance between the neighbor trackpoints with the haversine formula to compute the distances in bash with bc and drop the 2nd if it is more than 15km away from the 1st position.
If there are multiple occurencies of the same (wrong) spot, i delete all of them. Afterwards i restart the script for a new run (by a while loop).
This needs A LOT of time, but the maps are much better after that...
... as long as there is no real distance in the track which has really no position for a longer distance. That happens for instance if you going on a seagoing ferry to ride to Bornholm or a simple flight. :-(
So - been there, done that. No solution so far. Project stopped.
But you might be interested anyhow, so take a look.
We used a portable wifi2gsm router in denmark (2016 had still expensive roaming fees, we bought one dk sim card and used it together in the car). One of the offtrack locations is the sheep farm of a friend where i used the same device the first time (years before 2016).
heatmap.py creates the map section on its own, so it is smaller.
The script is still dangerous: load-trackfile.sh - have a copy of your precious data before tinkering around with it!
Old, but paid :-) heatmap.py
Needs this patch to work with recent version of osmviz
Makes nice heatmaps, if you do not use google location data but real gps data.
OSMBASE=https://tiles.wmflabs.org/bw-mapnik
./heatmap.py -B 0.35 -o 2021-11.png --osm --osm_base $OSMBASE -H 2400 --margin 50 --decay 0.90 ~/gd_local/tracks/2021/2021-11
If you combine real and google guessed data, it might look like this one (click to enlarge):
(from leben-auf-dem-boden - 2021/12/05 - november eher ruhig)
I was in fact on the island Amrum, but not on Föhr or Sylt, but my mobile connected to gsm antennas there; and i used the train to Schleswig and from Husum, and the train used the railway as mapped, not the plot printed by the gsm connections. The trains from Cuxhaven and Lübeck did it better :-)
Heatmap.py uses OSM for the map layer, there are a lot of possible maps out there, take a look at OSM tile servers listing to compare and find the suitable for your needs.
I like https://tiles.wmflabs.org/bw-mapnik (bw, free - used by the example maps)
You need to mention the openstreetmap copyright if you using OSM map tiles! Find an easy solution to add the right copyright into your maps further down in the tools section.
If you get your google location history via [google takeout](https://takeout.google.com/}, use the kml export if you want to use the typical gis tools and not the json export (which is default).
This nice script converts kml into gpx.
google-location-history-to-gpx
The swiss army knife for location data. https://www.gpsbabel.org/ Used it to split the all.gpx file in daily files.
I use imagemagick to add the osm copyright - for the toner tiles from c.tile.stamen.com i like grey40 as font color, i use their (c) string and push it in to the right bottom corner with convert -font helvetica -fill grey40 -pointsize 30 -gravity SouthEast -draw "text 20,20 'Map tiles by Stamen Design, under CC BY 3.0 Data by OpenStreetMap, under ODbL '" tmp/2021-12-3.png tmp/2021-12-crights.png.
With OSMBASE=http://c.tile.stamen.com/toner the background brightness -B 0.15 shows clearly water but is not too bright, e.g. OSMBASE=http://c.tile.stamen.com/toner OUT=../tmp/2021-12.png ; ./heatmap.py -e 53.55,9.85,53.44,10.08 -B 0.15 -o $OUT --osm --osm_base $OSMBASE -H 2400 --margin 100 --decay 0.90 ~/gd_local/Meine\ Tracks/tracks/2021/2021-12-*
I like OSMBASE=https://tiles.wmflabs.org/bw-mapnik most, but it shows deep water in different shadings, this is at least in Hamburg not easy to view. I use a background brightness -B 0.35 for that tileserver - there is no (easy to reach) limit in tile loading.