- ▼ 2017 (9)
Sunday, June 25, 2017
basically finished modifying python scripts to batch download .fit files from the garmin IQ web site, extract the raw data, and convert to kml files to view on google earth.
just recently back to comparing extracted .fit data with gopro video files in order to tweak algorithm. a bit tedious as have over 700 waves. hope to update results in the next few weeks.
my focus is on speed and distance profiles vice minimums.
+ + + + + + +
some observations comparing the wave file generated by the app and what I get when processing the raw data (extracted from the tcx file). I only recently installed surf tracker 2 so will update once I get some rides in.
In applying the wave logic used by the app (e.g. minimum speed [9 kph], minimum ride time [6 seconds], at least speed [13 kph]) to the raw data there was
high correlation in # of waves (80 out of 82) and max speed of wave (72/82 exact match, 77/82 within .1 kph (rounding?)).
medium correlation for ride time. Same for 55 of 82 and 63 of 82 within 2 seconds
medium correlation for wave distance when ride time matched. -23% to +13% when time matched (52 waves)
Since, with exception of # of waves caught and max speed of those waves, there are significant discrepancies in the ride time and distance when post processing raw data i conclude they must be due to applying the algorithm in real time vice after the fact.
However, I did capture a significant number of waves on a gopro so was able to review the video in an attempt to validate the wave ride time. For the largest discrepancies (difference of more than 4 seconds) the raw data had zero's in the speed, lat and long for the difference in time. For example, raw data had valid data for ride time of 18 seconds, then zero's for an additional 15 seconds (most of which the watch was under water). The app recorded the ride time as 33 seconds so it may be the app logic continues to give credit for ride time until a valid (i.e. non zero) speed is below the minimum.
At the moment I do not have a hypothesis about the discrepancies in wave distance. The majority are shorter on the app than post processing (10 longer , 45 shorter when waver ride time was same)
I have updated my python code to capture the time zone and street location of each wave. time zone was necessary to correct to local time as tcx time data is gmt.