Vegetation shading is only one sided

[FlorianK13]

Houses are not shaded when vegetation is directly above them. I guess this comes from the fact that we only have shading in the positive direction of the normal vector, but I'm not sure if this is true.

[khdlr]

Will have a look, can you share the coordinates / simulation link for this building?

[FlorianK13]

Sure @khdlr: /simulation/11.649088711250243/48.1737942262813

[khdlr]

Alright, for that link, the simulation center is quite far from the actual building. Here's what I think is happening:

We have 2 types of vegetation geometries, surrounding and background. Only the surrounding vegetation geometry is used for the raytracing simulation, the background is just optical (similar to what we're doing with the buildings).

Visualization of the two "zones" of vegetation for our dear Lange Point (background vegetation in gray, surrounding in green):

In your example, the displayed vegetation is purely background vegetation, as the cutoff is calculated from the "simulation center" specified in the URL, not the actual house's coordinates.

Manually moving the simulation center closer to the building leads to a more correctly shaded building:

Suggested Fix: After re-centering the scene on a building geometry, we need to do all distance-based threshold calculations based on the building's center and not on the currently used, arbitrary "simulation center"

[khdlr]

Alternative hotfix: Just checked what would happen if we simulate with all the vegetation that was loaded (i.e. make all vegetation "green" according to the above colour scheme). This still seems to be acceptable in terms of simulation duration.

[FlorianK13]

I think we could #394 instead, because I assume this is where the bug originates?

[FlorianK13]

But based on your image, we should at least increase the radius for vegetation shading. Maybe this also solves #321

[khdlr]

Good catch, just checked, #321 is indeed the same issue:

I think #394 should definitely improve things, but it's also just a kind of band-aid as it doesn't solve the underlying problem: Currently, our system does not center the simulation on the building.

Another issue that arises in #321 is that we're currently working with hardcoded radius thresholds, which breaks for large buildings.