Visibility graphs have a huge number of polygon crossings

[ekrell]

I am trying to create a visibility graph from a shapefile, following the examples.
The example programs run fine, and I am able to get a correct path.
For my own shapefile, the visibility graph seem to never avoid crossing the polygons. So solution paths always cross a polygon if a polygon exists between them.

Here is my original map, plotting the polygons with matplotlitb.

For faster debugging, I made a reduced shapefile with only two polygons.
Here, I plotted all the edges in graph.visgraph.get_edges(). You can also see the solution path passing through the polygon.

Any idea what might cause this? Happy to share code if it helps.
The code starts by converting a raster to polygons with fiona, saving them as a shapefile, then following example 1. Not too experienced with geospatial data, so I could have made an error here. But the polygons load fine in QGIS.

[ekrell]

Just to add some extra details,
I remade the shapefile in QGIS in case I was doing something wrong with the python libs. I got the exact same result, so I think my shapefile is fine.

Can the fact that this is a very high res, zoomed in region (compared to the globe-trotting example scripts) affect it? The coordinates are just not very different from point to point --> float errors?

[ekrell]

Update: coordinates seem to be at play. I scaled each by 1000 and got something better: far fewer polygon intersections!

The graph:

The path:

[ekrell]

I was able to come up with a work-around. The coastlines shapefile worked and so did my own hand-drawn test file in QGIS worked. The only one that didn't work was a raster converted to a shapefile. So I pulled up my raster->shapefile in QGIS, redrew a less "blocky" shapefile on top (still complex with lots of points).

I input this new version and it had only about 5 polygon-crossings. But after rounding and normalizing the coordinates, there were no edge crossings.

Conclusion? The blockiness of a converted raster does not play well with the VG generation. Ultimately my pipeline will be automatically converting rasters to polygons for this, but I should be apply to smooth them a bit in the code.

[TaipanRex]

Yeah this is likely rounding issues when you have a large amount of vertices that are very close to each other. Reducing the number of vertices is the best fix, as well as scaling like you did.

[ekrell]

With more data, still seeing a number of crossings. I have done numerous trials with number of rounding digits, range of new scale, etc. Note the above only shows a sample of the edges.

More polygon smoothing might help, but they are already lower resolution than I would like. Specifically dealing with nearshore USVs.

Anything else I should try?

[ekrell]

Just for reference, here it is with much smoother polygons. Like you said, it is sensitive to shape complexity. I would like to work with higher resolution, but I can get by with this for now. Later I can look into tuning this part.

[SteynJanus]

Hi Ekrell. Did you already happen to find a nice solution to this? The final image seems to have no crossing but also has a high resolution. Was it just a story of finding the right scaling factor and smoothing factor? Also, did you scale it in pyvisgraph? Thanks in advance :)

[ekrell]

Hello SteynJanus,

Sure enough, my solution was just to apply a little bit of smoothing to the polygons.

You can see the full code here where I build the shapes (some very rough code made for a class project....):
https://github.com/ekrell/conch/blob/master/planners/vgplanner_build.py

And the exact line is here (101): shape_ = shape_.simplify(0.0005, preserve_topology=False)

This resolution turned out fine for my needs. Hope this helps :)

[eagerbeaver04]

Hello, may be you can help me. I installed this package today and tried to get shortest path. I already tried rounding and smoothing.

[TaipanRex]

@eagerbeaver04 this is likely due to floating point representation issues as your coordiate system uses quite large numbers. take a look at this documentation and try reducing COLIN_TOLERANCE in visible_vertices.py (assuming you have cloned the repo), or just make your coordinates smaller by dividing by 1000.

[eagerbeaver04]

Thank you very much!