Perspective Matching Tool Camera, Vray Camera Problems.

[rajdarge]

I have just used the "Perspective Matching Tool" in max 2014 to create a perspective matched standard Camera.
The actual quality of the match is pretty good.
If I try and use a Vray Camera to match the same image, the x/y/z (or just xy) alignment tools just spin out of control and the camera flips, and I basically lose control of the camera.
If I use a standard max Camera and then load a Vray camera into the scene, select the camera port, press P and then "ctrl-c" in that port, the vray camera gets translated, but the perspective is completely off.
If I change the Max camera to a targeted camera and then adjust the focus distance, again the viewport looks "wrong" and the illusion is destroyed.

So what has happened to the max camera, why is behaving in this strange fashion, can I convert it to a Vray camera. And why is the focal length way different to the EXIF data that I have from the .CR2 that I shot?
Raj

[AlexP]

The way Ive done this in the past is just continuing to use the max camera as the view, and then using exposure/DOF overrides to compensate for not using a vray physical cam.

[rajdarge]

I wanted to to try and match the DOF to the original Shot if I could, and the motion-blur....
Is this still possible to get "physically" correct DOF, and Moblur, and also I have a lens file for the lens that I used for the shot....

[AlexP]

I wanted to to try and match the DOF to the original Shot if I could, and the motion-blur....
Is this still possible to get "physically" correct DOF, and Moblur, and also I have a lens file for the lens that I used for the shot....

Personally, Ive never had the settings exactly match between 3D camera and real world camera. Even though the perspective etc seems to match visually, the settings are definitely different as you mention.
Would also like to know if this is possible

[kosso_olli]

Personally, Ive never had the settings exactly match between 3D camera and real world camera. Even though the perspective etc seems to match visually, the settings are definitely different as you mention.
Would also like to know if this is possible

I was told that some of these issues are related to the aperture width settings in the common tab of the render settings window. With a lot of cameras you have to adjust it from the default value of 36mm. We often do stills with backplates from Hasselblad cameras, so we have to change that value to 49 or 51 depending on the camera. After that, the readouts for FOV or focal length are matching quite close. There was a formula to calculate the aperture width settins, will see if I can find it.

[rajdarge]

I have a feeling that the perspective match tool is somehow distorting the camera with an anamorphic scale. I noticed that when I turn the standard camera into a target camera that the null object is no longer aligned with the line helper line that joins it to the camera.
I am using a full frame dslr, and not medium format, and it's just a standard 24-70 zoom: no fisheye or weired tilt shifting.

[kosso_olli]

May I ask which full frame DSLR it is?

[rajdarge]

Yeah sure. It's a Canon 5d MKIII.

[svetlozar.draganov]

It seems that this tool works very strange even on Max native cameras, and it doesn't work well on Vray ones.
Have you tried Camera Match tool instead Perspective Match one?

[rajdarge]

Yes, I've tried using Camera match tool, but you need a model of known dimensions in order to use it at all. Its just a bit clunky.
After a lot more experimenting I think I may have the source of the problem. I just use the YZ as the object I was getting the angles from was not quite orthogonal to the other lines I was using. After removing the aberrant dimension (in this case x) the camera match was nigh on perfect and it even reflected true focal length (23.3) from the EXIF (calculated focal length for 35mm equivalent). So the fault was probably mine, as my lines were causing distortion. It actually tells me how powerful the tool is to cope with anamorphic distortion. Maybe one day it will work with Vray Camera.
After I fixed the perspective, I then was able to copy a Vray camera to duplicate the perspective...yay!

[AlexP]

Interesting, Ive not tried using just 2 axes, must give it a go

[kosso_olli]

A quick example of the things I was talking about in own of my previous posts, even though it was not the case with the problem here.

As you can see, with the correct aperture width for the Hasselblad H3D II, I get the correct readout for the focal length. When I set the aperture width to the default of 36mm, I get 26,25 as focal length. The perspective still matches, but the numbers are wrong.

[rajdarge]

Because the Perspective Matching tool can distort the camera a lot, I guess you have to be sure that the Perpendicular lines and parallel lines are actually that. I ran into trouble when I used lines I wasn't sure about, but I thought it would help having a bit more info.
I started with:


This was just a lighting test shot that I did in the studio, I used it because it had some of what I wanted and a known object (the filing cabinet and the edge of my backdrop). I would have normally culled this shot from the final selection but it seemed suitable for testing.

I then cut out the figure, and left in the guide lines. I made the mistake of using the horizontal lines of the filing cabinet when I wasn't sure if the cabinet was lined up properly. When I realised that I could depend on the Cut in the concrete floor of my studio and the verticle line of the filing cabinet and the edge of the backdrop things worked better. I imported this into 3ds max in the way described. I used it to eyeball the perspective match.


After this was done I compared it to the exif (from EXIFTOOL) that showed me this:
Focal Length : 24.0 mm (35 mm equivalent: 23.3 mm)
Hyperfocal Distance : 1.86 m
Lens : 24.0 - 70.0 mm (35 mm equivalent: 23.3 - 68.0 mm)
and also:
Focus Distance Upper : 1.6 m
Focus Distance Lower : 1.38 m
from these numbers I could then set up the distance to the subject and get the scale right.
When I looked at the focal length in max it was precisely 23.3 !

this was my final result.


In your example maybe you should ignore the painted lines, or at least some of them, and use the building lines (like the concrete) as most likely they are perpendicular.
Hope you get better results. For a first test I was pretty happy. I now need to find something that I can include in my scenes with models so I can transpose them. I have the luxury of knowing precisely how I shot the image and what light I used as I have a studio. I have yet to experiment with the mirrored ball shots in my studio though.
raj