Clustering GPS points based on distance between points?

Hello everyone!

Can some one of you how can i cluster GPS points. Let say i have a ploted on a map some GPS points and i want to cluster them based on the distance between each other.

[Paola Ceccon]

Hi, Have you resolve this problem? I have the same... I would appreciate your help

Answer 1

[Chad Greene]

If your data are in lat/lon coordinates, you'll need to convert them to some reasonable x/y coordinates, then you can use ipdm to calculate the distance from each point to every other point. Or you could identify centers of clusters and use hypot to calculate the distance between each point and each cluster center.

Either way, make sure you convert lat/lon to x/y because a degree of latitude does not equal a degree of longitude, so hypot(delta_lat,delta_lon) would not make much sense at all.

Answer 2

[Sean de Wolski]

Given a Geopoint P, which has an elevation property (in some unit) you can calculate the distance between the points using a referenceEllipsoid such as the wgs84 one:

% Build an ellipsoid with feet
wgs84 = wgs84Ellipsoid('ft');

% Distance in components
[dx, dy, dz] = ecefOffset(wgs84,...
  P.Latitude(1:end-1), P.Longitude(1:end-1), P.ElevationFt(1:end-1),...
  P.Latitude(2:end), P.Longitude(2:end), P.ElevationFt(2:end));

% Distance
dists = hypot(hypot(dx,dy),dz);

Note that the elevation and ellipsoid are both in feet for this example.

From here, if you want to cluster, you can use any standard method whether it's just a histogram or kmeans etc.

[DuckDuck]

Ok , i created the distance matrix between all points, using the distance and deg2km functions in matlab. But how do i cluster them based on position?

[Sean de Wolski]

I'd start here:

http://www.mathworks.com/help/releases/R2014b/stats/cluster-analysis-1-1.html

Actually, I'd probably start with kmeans because it's quick and easy.