[Closed] which facets are visible?

this isn’t nearly as easy as you were hoping, I’m gonna guess

For faces facing towards/away from the camera it’s easy enough.

1. Get the camera’s direction ($.dir) this is a vector.
2. Get the Normal of the face (getFaceNormal <mesh> i) this, too, is a vector.
3. Use ‘dot <vector1> <vector2>’ to test whether the face is facing away from (negative value) or towards (positive value) the camera.

For faces being occluded or not – welcome to a world of hurt that starts with this thread:
http://forums.cgsociety.org/showthread.php?f=98&t=295257
Basically, for each face you would have to check if any of its vertices are in front of all the other faces in the mesh; if a vertex is in front of another face, then some part of the face itself should be in front as well. The ‘in front of another face’ part is the annoying bit. If all the vertices of the other face are further away from the camera than the vertex you’re checking, then the vertex you’re checking is in front (similarly, if they’re all closer, then the vertex is behind). If one or two of the vertices of the other face are -closer-, however, you’ll need to do tricky math.

Dare we ask what you’re trying to do, just in case there’s a simpler solution?

Hi Uri and welcome,
here is a function to just look for faces orientation, it doesn’t consider occlusion, but it’s a starting point. It works with Editable Poly and Editable Mesh.

-- getFrontFaces accepts Editable Poly and Editable Mesh objects and returns
-- a bitArray containing Camera-Facing faces index

function getFrontFaces obj =
(
    -- general threshold
    local fThresh = 1e-4

    -- initialize face index storage
    local baFrontFaces = #{}

    -- initialize variables
    local iNumFaces = 0
    
    local p3FaceCenter = [0,0,0]
    local p3FaceNormal = [0,0,0]

    local p3CameraToFaceVect = [0,0,0]
    local fDotViewAngle = 0.0

    -- get current camera e viewport data
    local p3CameraPos = (inverse(getViewTM())).row4
    local p3CameraViewDir = -(inverse(getViewTM())).row3
    local bIsPerspective = viewport.IsPerspView()

    -- if object is an Editable Poly
    if ((classOf obj) == Editable_Poly) then
    (
        -- get current number of faces in object
        iNumFaces = polyOp.getNumFaces obj

        -- cycle through faces 
        for i = 1 to iNumFaces do
        (
            -- get face center and normal vector
            p3FaceCenter = polyOp.getFaceCenter obj i
            p3FaceNormal = polyOp.getFaceNormal obj i

            if (bIsPerspective == true) then -- view is perspective
            (
                -- get vector from camera to face center
                p3CameraToFaceVect = p3FaceCenter - p3CameraPos
                -- get dot between vector form camera and face normal
                fDotViewAngle = dot p3CameraToFaceVect p3FaceNormal
            )
            else -- view is orthographic
            (
                -- get dot between camera view direction and face normal
                fDotViewAngle = dot p3CameraViewDir p3FaceNormal
            )

            -- if face is not looking backward, collect it
            if (fDotViewAngle < -fThresh) then ( baFrontFaces[i] = true )
        )
    )
    else if ((classOf obj) == Editable_Mesh) then
    (
        iNumFaces = meshOp.getNumFaces obj

        for i = 1 to iNumFaces do
        (
            p3FaceCenter = meshOp.getFaceCenter obj i
            p3FaceNormal = getFaceNormal obj i

            if (bIsPerspective == true) then
            (
                p3CameraToFaceVect = p3FaceCenter - p3CameraPos
                fDotViewAngle = dot p3CameraToFaceVect p3FaceNormal
            )
            else
            (
                fDotViewAngle = dot p3CameraViewDir p3FaceNormal
            )

            if (fDotViewAngle < -fThresh) then ( baFrontFaces[i] = true )
        )
    )
    else
    (
        throw "Wrong input in function getFrontFaces()"
    )
    return baFrontFaces
)

• Enrico

Dare we ask what you’re trying to do, just in case there’s a simpler solution?

basically i need to know which faces are visible at a specific camera/object orientation ,and then color the visible faces according to temperature (temperature calculations done in a different program), coloring everything and then viewing through a camera view is NOT sufficient

i dont really care if the occlusion is partial (i’ll just say it is occluded)

while reading some help files i saw “deselectHiddenFaces <mesh>”
can i somehow select all faces ,then remove from selection the hidden ones and get a list of the remaining selected faces?
Mesh Face Methods[left]
[/left]

thank you so much for your replies

Hi Uri,
I just finished to write the whole function. It takes into account face direction and occlusion. It’s still a bit messy, not fast, and not commented, but seems to work well enough. I’ll try to refine it in the near future. It’s too long to post in this thread, you can find it here under “Snippets” -> “Get Front Faces”. Mind that you also need the function “getPolyFromMeshFace” under “Editable Poly data from underlying Mesh”. I hope it helps.

• Enrico

Thank you so much:applause:

i got it to work almost perfectly (including UI ) but i see that the selection is not perfect,
there are a few faces that are definitely not occluded and their normals are exactly like their neighbors which dont get selected,
any possible reasons ?

i converted obj to mesh

thanks

Hi Uri,
the function I posted is not perfect, it doesn’t take into account some cases. The problem you’re reporting seems to be tied to threshold accuracy. I’ve been working a bit on the whole function today and will post an update as soon as it is ready. Unfortunately it is still slow because heavily relies on the MaxScript IntersectRayEx function, proven to be quite slow.

Hang in there

• Enrico

thanks,
this part:

 if (fDotViewAngle < fThresh) then 
  			( 
  			baFrontFaces[i] = true 
  			)

returns true for those faces , so i guess they are later made false,

i think that in those faces a vertex or more ARE occluded and it removes the hole face

is there a way to check the face center instead of its vertices (should also be faster no?)

i did this:

		for i = 1 to iNumFaces do
 		(
 				p3FaceCenter = meshOp.getFaceCenter obj i
 				p3CameraToVertDir = normalize(p3FaceCenter - p3CameraPos)
 			
 				rTest = ray p3CameraPos p3CameraToVertDir
 				aResult = intersectRayEx obj rTest
 	
 				if (aResult != undefined) then
 				(
 					if ((posMatch p3FaceCenter aResult[1].pos fThresh) == false) then
 					(
 						if (distance p3CameraPos aResult[1].pos) < (distance p3CameraPos p3FaceCenter ) then
 						(
  							baFrontFaces[i] = false
 						)
 					)
 				)
 	
 				p3RayPos = p3FaceCenter + p3CameraToVertDir * fThresh
 	
 				aIterator = #(true)
 				for dummy in aIterator do
 				(
 					rTest = ray p3RayPos p3CameraToVertDir
 					aResult = intersectRayEx obj rTest
 	
 					if (aResult != undefined) then
 					(
 						baFrontFaces -= #{aResult[2]}
 	
 						p3RayPos = aResult[1].pos + p3CameraToVertDir * fThresh
 						append aIterator true
 					)
 				)
 	
 				rTest = ray (p3FaceCenter - p3CameraToVertDir * fThresh) -p3CameraToVertDir
 				rResult = intersectRay obj rTest
 				if (rResult != undefined) then
 				(
 					baFrontFaces[i] = false
 
 				)
 			)
 			setFaceSelection obj baFrontFaces
 		)

i think it does what i wanted…
are there errors in this code?
for instance i think this: baFrontFaces -= #{aResult[2]}
is probably an error i just have no idea what to put there