Path tracing producing awkward artifacts - Forum Open Cascade Technology

Pawel
Wed, 02/01/2023 - 16:23

Forums:

Hello,

I have defined a scene and use it to render images using path tracing with OCCT770.
The components of the scene are:
- a CAD modell (stud)
- light source (V3d_PositionalLight)
- a grid that is placed in front of the light source (stripes)
- a camera

The generated stripes do not seem to be right and obviously depend on the angle between the light and the face normal (see attached image).

It looks to me like OCCT bug. Can anyone confirm that?
Or maybe there is something I'm doing wrong defining the scene?

Thanks
Pawel

Attachments:

Cylinder.PNG

gkv311 n
Wed, 02/01/2023 - 17:20

Is it an orthographic protection? How it looks with perspective camera?

Pawel
Mon, 02/06/2023 - 14:27

This is a perspective view. But it doesn't matter if the viewer is in orthographic or perspective mode. The effect is the same.
Image (orthographic projection) attached.

Attachments:

Cylinder_orthographic.PNG

gkv311 n
Tue, 02/07/2023 - 12:45

Would be nice if you could share a Draw Harness script or C++ code with the model for reproducing this scenario to give some comments why this might happen.

Pawel
Tue, 02/07/2023 - 18:19

OK, I'll try to create a corresponding Draw Harness script and will upload it here.

Pawel
Tue, 02/07/2023 - 19:20

Ok, I have written the following script (see below). I'm not sure if it reproduces my effect in 100%. But it can be clearly seen that at the border of the cylindrical face and the plane the light does not traverse correctly.

I will try to create some more examples tomorrow.

pload ALL
vinit
pcylinder c1 25 50
vdisplay c1
vsetdispmode c1 1
vsetmaterial c1 plastic
vsetcolor c1 white
box b1 50 5 -200 1 50 400
vdisplay b1
vsetdispmode b1 1
box b2 50 -5 -200 1 -50 400
vdisplay b2
vsetdispmode b2 1
vfit
vraytrace 1
vrenderparams -shadows 1 -reflections 1 -fsaa 1
vlight redlight -type POSITIONAL -headlight 0 -pos 100 0 100 -color RED -intensity 10

Attachments:

draw2.PNG

Pawel
Wed, 02/08/2023 - 14:15

I was able to create a script that shows the problem.
Image attached.

pload ALL
vinit

pcylinder c0 50 5
vdisplay c0
vsetdispmode c0 1
vsetmaterial c0 plastic
vsetcolor c0 white

pcylinder c1 25 50
vdisplay c1
vsetdispmode c1 1
vsetmaterial c1 plastic
vsetcolor c1 white

box b1 1000 -0.5 -200 1 1 400
vdisplay b1
vsetdispmode b1 1

box b2 1000 -1.5 -200 1 -1 400
vdisplay b2
vsetdispmode b2 1

box b3 1000 1.5 -200 1 1 400
vdisplay b3
vsetdispmode b3 1

vfit
vraytrace 1
vrenderparams -shadows 1 -reflections 1 -fsaa 1
vlight redlight -type POSITIONAL -headlight 0 -pos 1025 0 200 -color RED -intensity 10

Attachments:

draw3.PNG

Pawel
Wed, 02/08/2023 - 15:02

just one more thing: I ran all the scripts with OCCT 770 but the effect should be the same in lower versions also.

gkv311 n
Wed, 02/08/2023 - 17:21

Looks like accumulated error of single precision floating point math. As larger distance of the light - the harder effect.

pload MODELING VISUALIZATION
vinit View1
vclear
vraytrace 1
vrenderparams -shadows 1
vlight redlight -type POSITIONAL -headlight 0 -pos 25 0 200 -color RED -intensity 10 -display

pcylinder c0 50 5
pcylinder c1 25 50
vdisplay  c0 c1 -dispMode 1
vaspects  c0 c1 -material plastic -color white
vfit

box b1 0 -0.5 -200 1  1 400
box b2 0 -1.5 -200 1 -1 400
box b3 0  1.5 -200 1  1 400
vdisplay b1 b2 b3 -dispMode 1

proc testdist {d} {
  vlocation b1 -setLocation $d 0 0
  vlocation b2 -setLocation $d 0 0
  vlocation b3 -setLocation $d 0 0
  vlight redlight -pos [expr 25+$d] 0 200
}

testdist 100
#testdist 1000

Pawel
Thu, 02/09/2023 - 12:09

Would there be a work-around for this problem?