Cineversity

Hi everybody,

Just a little thing that I would like to touch on to make my workflow better:

My work requires me to take photos of real objects and reproject them onto models of (matching) C4D objects, and then bake them all into a single UV-mapped texture for display in a little proprietary OpenGL viewer. That way, all of the little details, decals, and lighting are from the photographs rather than needing to be modeled, have materials, etc.

Of course, this requires me aligning my camera view of the model to the photographs of the real object. I would like to get these matches as close as possible, to avoid manual fiddling with the UVs.

Here’s the thing: to get the best projections (see another thread about this that I started), I’m constrained to shoot my photos aligned to the global axes, not oblique to them. That means that, in the Camera Calibrator tag, it’s very hard to get vanishing points. Also, the Camera Calibrator tag doesn’t like it when you try to calibrate two cameras on opposite sides of an object.

Does anyone have any tips for doing this matching process another way? I currently set the focal length and sensor size to that of my real camera, and then point it at a null, so that I can move the center and the camera independently. Then, I basically fool around with the positions until I’m satisfied. But is there a quicker, better, more systematic way to do this?

Any help would be greatly appreciated! Or, if I haven’t described it clearly enough, I would be happy to expand on what I’m doing and what I need.

Best wishes,
Eric

Hi Eric,

In an upcoming trainings series I discuss the problems when objects have no option for vanishing points, and the missing lines pointing to those. I call it the “5 foot cube for the pocket”.

However, the simplest way is certainly given by placing something in the background, which gives you those lines. If you like to continue with the Camera Calibrator Tag, perhaps you build a cube out of little wood sticks, which should be larger than your object(s). A half cube might work already. (Half—to cast no shadow on the object) In a studio situation we normally place some C-Stands in the image (green-screen), to get those lines. Two cardboards with checker box-pattern—90º to each other—might do a better job in your case.

NOTE: the Camera Calibrator Tag is only needed if you have no idea where the camera was in conjunction to the object in the image, etc.

In your case, you have objects that (so far I understand) are not that large in size. If you do a good “shooting-survey”, you can set up the camera in C4D very fast. In your case, with the 90º shooting to get six cameras, I assume you have the camera or the object always vertical/horizontal. Which makes it pretty simple. The only problem that you might have, is based on the use of a zoom lens (if focal length varies), instead of a fixed focal length: a prime lens. Will say, stay with one focal length, and measure this lens with the camera calibrator, if there is no other option. Once you know exactly what each lens does, and the you have the distance between camera and object, it should be simple to set everything up. Measure the distance from the camera body, there should be an o with I in the middle, that is the sensor (film) position.

Again, field of view, distance and orientation of the camera is all the camera calibrator tries to figure out. If you know those data, use these. Real world measurements are better from my point of view.

All what you do with the Camera-Calibrator Tag is—to re-create the scene that you had while shooting, nothing more.

Shoot with this lens a lens grid! Get the lens to know, and how to get un-distorted images from it. More later about that theme*. (The *theme is “Photography for teh 3D Artist”, and I’m half through it. It a long series, 200 parts, but really researched and in depth)

All the best

Sassi