1

I have an polar grid array in numpy with values in range from -31.5 till 80. We want to detect the clusters based on neighbour values.

Something like this:

enter image description here

And this is a good example for our polar grid structure (don't look at the arrows):

enter image description here

Now the point, what is the best way to use dbscan for this to get the number of detected clusters ?

Do you also need a printout of the numpy array ?

Thank you!

user3408380
  • 375
  • 3
  • 5
  • 22
  • For those not familiar with it, [DBSCAN](http://en.wikipedia.org/wiki/DBSCAN) is a machine learning algorithm. To the OP: Are you sure that you want something like DBSCAN? It's for randomly distributed points, not information on a grid. Traditional image analysis is likely to be much faster in this case (in my opinion, anyway). – Joe Kington May 05 '14 at 19:36
  • Thank you, I got this from a study. Check this out on page 3: http://www.vaisala.com/Vaisala%20Documents/Scientific%20papers/7.Matthews,%20Trostel.pdf – user3408380 May 05 '14 at 19:44
  • Note. I'm also open for other solutions is you think that the image analysis is better! – user3408380 May 05 '14 at 19:53
  • Are your grid elements of equal-area so that the values can be compared directly? Or does some sort of Jacobian come in to convert from counts to probability density? – Raman Shah May 05 '14 at 22:21
  • @user3408380 - Well, the image processing version is basically identical to this answer: http://stackoverflow.com/questions/4087919/how-can-i-improve-my-paw-detection/4092160#4092160 (caveat, plugging one of my own answers there). It's basically what Jamie suggested for one of your earlier questions. However, blurring and filling holes will give you much cleaner regions. You can also automatically get tracking out of it if you put in a 3D array (i.e. stack each map into a 3D array). There are a couple of other caveats, but I don't have time for a complete answer right now. Hope it helps! – Joe Kington May 06 '14 at 01:23
  • Thank you! The grid elements are equal area! – user3408380 May 06 '14 at 07:39

0 Answers0