Take the 2-minute tour ×
Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. It's 100% free, no registration required.

This winter I am planning to track my downhill skiing/snowboarding using a GPS. Most of my riding will occur at the same resort. I would like to be able to create a sort of "heatmap" that shows the amount of runs that I have made in a given area. As I add more and more GPS traces to my database, my goal would be to see a sort of linear heatmap of the most traveled areas. Given the nature of downhill skiing, you would expect the uphill chairlift lines will be the "hottest" because it will be the only places visited over-and-over again.

Given that 1) my track will not be the same every time and 2) the area covered by following one "run" may be a few hundred feet wide, what might be the best way to analyze this "linear" data to create a sort of heatmap? My thoughts were to buffer the lines, then intersect the polys to get a sort of Venn Diagram thing going. My preference is to use open-source technologies. I've got QGIS and PostGIS loaded and available.

UPDATE: In regards to @blah238's response, I was thinking of something that might be able to "collect" the number of passes ("runs") through an area, and then symbolize by the count. Conceptually, this would be similar to ArcGIS "Collect Events" (but for lines, not points) or Collapse Dual Lines To Centerline (but for multiple line in roughly the same area).

A more visual example of a similar concept might be a traffic-flow map, where highly-congested areas would equate to "highly-traveled" ski runs/areas:

Google Traffic Map

I've read the following questions that may give some ideas, but they don't really address what I am trying to accomplish:

Clustering Trajectories (GPS data of (x,y) points) and Mining the data

Managing error with GPS routes (theoretical framework?)

share|improve this question
    
Assuming that you would be using GPX data, it should just take lat/long readings at intervals, so it should be already in point features. Would you not be able to then use a heatmap on that? Or are you not using GPX compatable device?? –  RomaH Nov 26 '12 at 20:55
    
Ah, er. I misread what you were wanting.. Nevermind. –  RomaH Nov 26 '12 at 21:51
    
@RomaH, actually, you very well may be onto something with your idea of using the points to generate a heatmap instead of using the lines. I had not considered using the source points before, so this is something I will definitely investigate. –  RyanDalton Nov 27 '12 at 15:18
    
You are dealing with a field, where as a street has well defined edges. If you want to do a line model, you could do several poly overlays, joins, and use the point2one plugin to make a line map with inherited values. But then you need to decide how far away does one need to ski from in order to justify a new line feature. Personally for me I would be happy with a heatmap solution, but again, I am not aware of your needs. –  RomaH Nov 27 '12 at 17:16

6 Answers 6

Here is a good tutorial for doing exactly that using MapBox and TileMill:

Screenshot

share|improve this answer
1  
Agreed, this concept is almost identical, and could easily be done in any GIS. But this is basically suggesting "just make all of your tracks red with a light green buffer" and they will stand out on the map? For visual effect, I will definitely work. I guess I was originally thinking of trying for something more "analytical" that could show "counts" of "runs" in this example. I'll keep mulling it over. –  RyanDalton Nov 27 '12 at 0:16
    
@blah238 I was looking similar example...thanks a lot !! –  Sunil Nov 27 '12 at 5:09
1  
@blah238, After thinking on it a little longer, you provided a very valid answer to my initial question. I should have been more clear on my thoughts. Thanks for your answer, and I have updated my question to more accurately reflect what I was trying to achieve. I appreciate the help. –  RyanDalton Nov 27 '12 at 15:20

I've done a bit of work on this in GeoTools/GeoServer by extending the Heatmap Rendering Transformation to support geometries other than points.

Pizza deliveries heatmap

It's not finished yet, but you can get the feature branch from my repository on GitHub.

The screenshot is of GPS tracks from when I worked as a pizza delivery driver.

share|improve this answer
    
It's still not outputting vector geometries, but it is collecting overlap to give a higher value. –  smithkm Nov 27 '12 at 17:56

here is my simple approach:

  1. create a new map in umap: http://umap.openstreetmap.fr/en
  2. click Import Data a select all the gpx files you have and upload them into map (you can import all of them at once)
  3. enter Edit map settings > Default properties, choose opacity 0.25, weight 10.

The three steps above will take 5 minutes and here is the result: enter image description here

share|improve this answer

Create a graticular gridwork, convert gpsdata to shapefiles, polygonize the lines, extract x,y point data from tables, make a surface density, or heat map using standard practice at this point. You could then merge with linear vector data to get pixel values in a raster display or for further numerical processing.

share|improve this answer

Most people (including myself) look at a GPX file and think of it as a sequence of positions (points), while it can be perfectly seen as a sequence of line segments instead.

Your problem formulation involves a "heat map" so that you can inspect visually your trajectories, so there is a rather simple way to skip the complex numbercrunching part by delegating it directly to a plotting engine.

When you say "buffer the lines", that would be line thickness. When you say "collect the passes" that would mean plotting over and over with transparent lines, thus "building up" more and more color.

So, I suggest you to plot each trajectory as a single polyline with thickness enough that there is some overlap (it depends on zoom, so you'll want to set the thickness in map distance instead of pixels), and with a fairly low opacity, say, 0.05, in black.

It is important not to plot each segment, because if you do so, their extremities will overlap, doubling their opacity and creating a "dot" of stronger color.

In the end, you'll get a grayscale image upon which you can apply the colormap of your choice.

I would probably do it using Python/Cairo, but Python/Matplotlib would do, and Html/Canvas or Html/SVG (or programmatic SVG) certainly would do.

The end result would depend only on resolution of your output media.

share|improve this answer

I'm aware that this is quite an old question, but as this post is one of the top search results for this topic, I thought I'd post my workflow for producing polyline heat maps in ArcGIS, as there is currently no answer for ArcGIS in this post.

This technique requires the polylines to be coincident, i.e. exactly overlaying each other. If your data does not match this criteria it is easily fixed using Integrate (if your lines are dense with points, running Simplify Line first, will produce better results but may reduce line accuracy). Once your data is prepared follow these steps:

  1. Run the Split Line At Vertices tool.

  2. Create two additional attribute fields, one will be an identifying field "ID" and another to count coincident line segments "Count". "ID" should be text, "Count" either a Short or Long integer depending on how dense your dataset is.

  3. Using the Field Calculator populate the "ID" field, ensuring the expression type is set to Python, the "ID" needs to be unique to the line segment but the same as coincident lines. The following expression is suitable:

    ID = "{0}{1}{2}{3}".format(!Shape.firstpoint.x!, !Shape.firstpoint.y!, !Shape.lastpoint.x!, !Shape.lastpoint.y!)
    

    It is possible to use other Geometry objects to create a unique identifier but this should suffice.

  4. Next to populate the "Count" field, run the following code in the Python window, changing the appropriate variables:

    import arcpy
    
    feature_class = "Polylines_FeatureClass"
    id_field = ["ID"]
    count_fields = ["ID", "Count"]
    
    id_list = []
    
    with arcpy.da.SearchCursor(feature_class, id_field) as cursor:
        for row in cursor:
            id_list.append(row[0])
    
    with arcpy.da.UpdateCursor(feature_class, count_fields) as cursor:
        for row in cursor:
            row[1] = id_list.count(row[0])
            cursor.updateRow(row)
    

    While adequate, this technique is definitely not optimised for large datasets, it's more of a guide on the workflow.

  5. Lastly run the Delete Identical tool, selecting the "ID" field for comparison. Warning this tool has no output and modifies the input data in place!

This process results in coincident lines being resolved to a single line, with the attribute "Count" representing how many lines were coincident, "Count" can then be used to represent the density through symbology. Whilst this process is more convoluted than overlaying semi-transparent lines, because the 'density' is quantified by "Count" other statistical techniques can now be applied, or just visualised using multiple-colour colour ramps which is not possible using the overlay technique.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.