Map matching and distance

rgugliel commented 5 years ago

Hello, I'm facing a result from a match query that I don't fully understand.

The context is the following: I have a GPS trace in Paris of around 1700, that comes from the following journey: https://www.google.com/maps/dir/48.870856,+2.342875/48.857898,2.307158/@48.8664598,2.3186734,15.25z/data=!4m17!4m16!1m13!2m2!1d2.342875!2d48.870856!3m4!1m2!1d2.3104721!2d48.862324!3s0x47e66fd9f701a3eb:0x24b6aa8a6b668b71!3m4!1m2!1d2.307392!2d48.8560377!3s0x47e67027704214cb:0xd63a6bcef681d1a6!1m0!3e2

I query osrm-backend 5.20, to get the match and compute the total distance of the route (I sum the distances over all the matchings)

If I call the match service with all the points, I get a distance of around 1000m. On the other hand, if I filter the points to ensure that they are not too close (as advised in the article https://infolab.usc.edu/csci587/Fall2016/papers/Hidden%20Markov%20Map%20Matching%20Through%20Noise%20and%20Sparseness.pdf ), I get a distance of 4364m, which is close to what Google gives.

What I don't understand is the following:

Why is the difference so big?
Why is the distance with more points smaller (I would have expected the opposite)?

Note: I draw on a map the path returned by OSRM. In both cases, they are very similar so it seems that the path is correct but the distance computation may have an issue.

What did I miss?

Thanks a lot.