Segment matcher:

Demo for my region (Algarve, Portugal)

https://segmentmatcher.appyou.pro

N.b. Load routes in the Algarve for this to be able to detect any segments.

If you wish to run your own version, you can add segments for your own region pulled from the Strava API as per the instructions bellow.

How to run:

• Navigate to the folder containing these files,

• Run node in the CLI (tested on Ubuntu only),

• Require or import the app:

  let gpxExt = require('./main');

• Add a segment (this was used to populate the app with actual segments from Strava) Pass the file path and the option "true":

  gpxExt('./testSegment.gpx', true);

• Load a GPX track and test for segments:

  gpxExt('./test.gpx');

Want to add segments for a given area from Strava?

Require or import, n.b. You may need to add an API key, https://www.strava.com/settings/api these expire quickly, hence mine was committed, it was not done in error.

let stravaSegments = require('fetchSegments.js');

Where swpt is the southwest point, and nept is the north-east point of the area you want segments for. Large areas are divided up into small 0.05 degree, by 0.05 areas.

stravaSegments(swpt, nept);

Nomenclature

• Segment: A JSON file containing an array of x,y,z points in positive space.
• GPX track: A GPX file of a track to find segments on.

How it works

Segments are stored with a unique name (a hash typically) in the file system, currently they are simply dumped in the segments folder, but a further implementation would follow the structure of an R-tree. This branches are made whenever a particular branch has 9 points, rather crudely by dividing the branch in two: A better way might be to use hierarchical clustering. The R-tree structure is used to store points and segment names to speed up look-ups.

If required, a seed object for the segmentRTree file is as follows:

{"minX":0,"minY":0,"maxX":0,"maxY":0,"branch":1,"leaves":[]}

Once a GPX track is loaded to find segments (this could easily be made interactive, as part of a routing application), a search is conducted for each point with an error of 25 m (radius).

TODO: Check following point to see if they provide a closer hit.

If a segment with the corresponding start point is found, the same GPX track is checked to see if the last point also has a point closely corresponding to the end point.

If the segment does have a closely corresponding start and finish point in the GPX track, the segment is checked against that portion of the GPX track for points within a given distance and then trimmed to return an actual matching portion of the track.

Why Node.js and C++?

I chose Node.js because it allows, for rapid prototyping and testing in my case. The design of the program using an R-tree to store segment start points also means that even with large data-sets, there is a very limited number of branches to descend.

It is however slow so I used C++ for two of the slower functions and will write others in C++.

In a production environment I would not develop my own solution as I have done here as a technical exercise, but use a database extension such as PostGIS for this purpose. Or if I was to do this in Node.js I would implement a preexisting solution such as RBush by Vladimir Agafonkin.

##Limitations: It does not catch errors in C++ as of now.

TODO:

• More tests, I wrote the minimum I need.
• Test accuracy
• Check duplicates segments allow slashes in names,
• Generate docs,
• Create a simple webview graphing the track, elevation and found segments (nearly done),
• Snap GPX to roadways, elevation correction...