Mesh Shadow Component Casts Backwards Shadows

Hello,

I'm putting together a script that calculates a percentage of a surface that is shaded at various times during the day. The backbone of this script is the mesh component.

However, I've run into major problems with the mesh shadow component becasue it cast shadows not only in the direction of the vector (away from the sun, like one would expect) but also opposite the direction of the vector (towards the sun, like one woule not expect). This means I cannot use the script I've written if there is any geometry beyond the plane that catches the shadows.

I've tried a few dozen ways of getting around this, like cutting / culling geometry that is behind the plane (the shadows are being cast upon), but what I've ended up realizing is that I just need a component that doesn't cast backwards shadows.

Attached are a few images that explain this problem. In this instance, the cube should cast a shadow on the plane, the sphere (which is opposite the plane from the "sun" should not).

mesh_shadow_script.jpg

mesh_shadow_persp.jpg

Replies to This Discussion

Permalink Reply by Jissi Choi on July 10, 2012 at 2:09am

Here is a solution.. using Dot product and Dispatch Component..

Attachments:

occlusion.gh, 7 KB

Permalink Reply by Piotr on July 10, 2012 at 5:04am

Hi Jissi

I have been recently struggling with the same problem (self casting shadows on a large objects) and I ended up with slicing breps.

I have tested your definition to see if this brings an answer but I must say with regret - it doesn't.

I've changed vector and a plane. Images attached.Looking forward to new try.

Piotr

Permalink Reply by Jissi Choi on July 10, 2012 at 8:08pm

I used Solid Differnce.. Check it..

Permalink Reply by Jissi Choi on July 10, 2012 at 8:11pm

Attached..

Attachments:

occlusion_trim.gh, 6 KB

Permalink Reply by Piotr on July 12, 2012 at 1:27am

Using fragment of your definition I have made mine, where you can cast shadows on multiple (flat) surface from multiple breps.

It works, but it's messy (for the first time I have used path mapper) so any cleaning & simplifying actions are welcomed.

Attachments:

brep test.3dm, 94 KB
shadow cast.gh, 14 KB

Permalink Reply by Jissi Choi on July 13, 2012 at 1:32am

Here it is..

I modified the ray vector to be parametric in front direction..

Attachments:

shadowcast_revised.gh, 16 KB

Permalink Reply by David Rutten on July 10, 2012 at 3:43am

Yeah I suppose that's a nasty side-effect of the code. I use the following logic to find those shadows:

Protected Overrides Sub SolveInstance(ByVal DA As Grasshopper.Kernel.IGH_DataAccess)
Dim mesh As Mesh = Nothing
Dim dir As Vector3d = Vector3d.Unset
Dim plane As Plane = plane.Unset

DA.GetData(0, mesh)
If (Not DA.GetData(1, dir)) Then Return
If (Not DA.GetData(2, plane)) Then Return

If (Not dir.IsValid) OrElse _
(Not plane.IsValid) OrElse _
(mesh Is Nothing) Then
Return
End If

Dim perp As New Plane(plane.Origin, dir)
Dim rc As Polyline() = mesh.GetOutlines(perp)

If (rc Is Nothing) OrElse (rc.Length = 0) Then
AddRuntimeMessage(GH_RuntimeMessageLevel.Error, _

"Shadows were not cast onto the plane")
Return
End If

For i As Int32 = 0 To rc.Length - 1
Dim pline As Polyline = rc(i)

For v As Int32 = 0 To pline.Count - 1
Dim ray As New Line(pline(v), dir)
Dim t As Double
If (Rhino.Geometry.Intersect.Intersection.LinePlane(ray, plane, t)) Then
pline(v) = ray.PointAt(t)
End If
Next
Next

DA.SetDataList(0, rc)
End Sub

If you find a good solution let me know and I'll adjust the algorithm.

David Rutten

david@mcneel.com

Poprad, Slovakia

Permalink Reply by Danny Boyes on July 10, 2012 at 6:09am

Hi David,

Maybe you can make something of this. Because of the null/paths bug it doesn't work all the time but I was trying to demonstrate it for one above one intersecting and one below.

Attachments:

Shadows.gh, 17 KB

Permalink Reply by Serge Neuman on July 10, 2012 at 4:42pm

If I understand well, the double loop is working on a projection, so I don't see how to immediately find out if the mesh is between the projection and the source of light or if it's the projection that is between the source of light and the mesh. The solutions I can see required to come back to the mesh. It can be done by projecting the point found (pline(v)) back on the mesh (MeshRay(mesh, ray) and then look if this point is in the same direction than the sun (dot product equal to the product of the length (a test distance between the two points = 0 or the dot product is > 0 sounds safer here )). But I'm afraid it would consume a lot of CPU...

Another solution is to have a look at mesh.GetOutlines(plane) to see if it can easily be written into a GetOutlines(plane, dir) and then consider only the points that are in one of half space defined by plane.

I'm sure there are better solutions. But it's very early in the morning in Europe right now and time for me to go to bed.

RSS