1

I am looking for a tool to analyze location data (long lat values from vehicle GPS available). The goal is to identify locations which are in close proximity (within a circle of 30-60ft) and then average values which are assigned to the different locations. Background is that it should be the same GPS values every time since the vehicle is visiting the same location but due to inaccuracies of the GPS the location slightly varies. I am trying to identify the location data which belongs to the same location. Any suggestions?

PS: First time user, let me know if you need more information.

Adding example data for clarification:

  1. long lat date reading
  2. 32...-81.. 1/19/2015 167
  3. 32...-81.. 1/20/2015 345
  4. 32...-81.. 1/22/2015 556
  5. 32...-81.. 1/25/2015 597
  6. 32...-81.. 1/28/2015 658
  7. 32...-81.. 1/30/2015 298

Is there software available which could group the locations in above example according to proximity (for example group all locations which are within 30-60ft) and then compute the average of the readings?

PolyGeo
  • 65,136
  • 29
  • 109
  • 338
Chris
  • 11
  • 3
  • Drop some decimals from the coordinate values? Is this supposed to be done in software or as a modification of data points or what? – Chris W Feb 26 '15 at 22:56
  • Yes looking for software, see sample data for clarification – Chris Feb 27 '15 at 16:10
  • I meant are you writing software that should do this internally, or trying to modify data using software; sounds like the latter. If looking to combine points within 30-60ft (9-18m), my suggestion was dropping decimals in your coordinates after 4 places since such a coordinate is accurate to about 11m. Basically what you want to do is establish a threshold/group points and then average gps points. We have a few questions here on the subject, and it could be done in several softwares. – Chris W Feb 27 '15 at 21:59
  • You might focus your search on GPS utility software. It's often free or less expensive than GIS software. A quick search came across http://factsfacts.com/geocachingsoft/AveragingCoordinates.htm but that may not handle processing large amounts of data at once. QGIS is free, and there may be a plugin out there that can do this specific task. Some of the questions in that search have comments or answers that talk about libraries that can do that function. I've got an idea of how to do it in ArcGIS, but if you don't already have that and this isn't personal, it wouldn't be cheap. – Chris W Feb 27 '15 at 22:09
  • I am now starting to understand your question. To confirm, I am trying to modify data using software. So far I actually worked in excel and based on this discussion found [http://www.neilson.co.za/k-means-cluster-analysis-in-microsoft-excel/], but the k means requires me to define the numbers of clusters, whereas I want the software to actually determine that for me. I like to dig more into your suggestion to drop decimals after for place. I do not believe it is highly complex as the clusters are easily defined. – Chris Mar 01 '15 at 17:00
  • For a purely mathematical or statistical approach, as opposed to GIS software specific, you may want to investigate over on Cross Validated, the stats stack exchange site. This subject (GPS) is also covered there in a few questions, and the topic of clustering in even more questions. – Chris W Mar 01 '15 at 22:09
  • Thank you, I think purely statistic might be a good approach given the topic – Chris Mar 02 '15 at 01:45

1 Answers1

1

Something like DBSCAN as an algorithm could help python implementation or DJ Cluster algorithm. Have a search for DJ cluster. A kdtree would also allow you to do a nearest neighbour lookup. Finally, there are functions, and spatial index's in PostGIS that you could use.

user965586
  • 159
  • 6