Multi-camera calibration

[HansWellens]

Hi there,

My question relates to the calibation of multiple DSLR's fixed stereo setup (12 fixed cameras). Are these/can these be calibrated separately?

Thnks,
Hans

[Andrea Alessi]

Hello Hans,

3DF Zephyr is completely autocalibrated but in case you want to calibrate them manually you can absolutely do so. You can find the calibration tool in the utilities menu. Camera calibrations can be imported then in the camera calibration page (in the new project wizard) by clicking "add new calibration" or via "modify calibration".

You may also import camera position constraints by checking the appropriate tick in the project wizard.

Cheers,

Andrea

[Modeler17]

Hi Hans and Andrea,

I am also considering a multiple DSLR fixed stereo setup. My use case would be for body scanning in order to make virtual fit models for clothing. I am thinking about an 8 camera setup arranged in 4 vertical towers of two cameras each. As a complete photogrammetry novice, in my 3rd day of Zephyr lite trial I have only one camera set up so far (and have posted some issues with my first project under another topic).. My other scanning experience is with scanners which rely on infrared depth data. The point clouds coming out of the IR scanner I am currently using (a three camera tower) are pretty messy . And there are no nice products, like Photoshop and Lightroom with which to work on raw infrared depth data. So the search is on for a way to get cleaner 3d images to start with.
Hence my interest in photogrammetry:
How would Zephyr best interpret the 8 camera configuration I am thinking of? For instance, would Zephyr be able to understand that there were 8 related circular sequences? How would one tell Zephyr that the data coming from the pair of cameras in each tower was related in a vertical way, while the data coming from tower to tower was to be understood as moving in a horizontal way?
My initial thought was to make separate projects from the data coming from each DSLR and then mesh the projects somehow.
But From your answer to Hans, Andrea, it sounds like I need to read more on position constraints. Is there an example or template for how to write a position constraint file for import?

[Scott Nebeker]

If you're looking to read more about any/all of this, I'd suggest 3DF Zephyr's User Manual

The specifics of the photography aspect are outlined starting on page 7 of the document.

The main point to remember from that document is [em]redundancy[/em]. I'm afraid that with a whole body scan, you may not achieve what you're wishing to accomplish with 8 cameras. A book for instance, photographed on a turntable (or lazy-susan) can be successfully recreated using what the software will consider two heights (similar to your two camera per height setup). This is accomplished by laying the book on the cover, photographing it with plenty of overlap as its spun in front of the camera. Then placing the book on it's back and repeating the process.

This takes a lot of skill and time and effort but is possible. On the human scale, I've found that if you're going for the accuracy you want, you'll need [em]at least[/em] three per tower, making 12 total cameras. The reason for this is that it's good to have what I've been naming things in my projects, the "level" shots, the "up shots," and the "down shots."

Example would be that you'll want clear coverage of the shoulders and top of head as well as the shots looking up to cover other parts of the body not seen from above (i.e. under the chest, hands, chin/jowl, under the ear, etc.). If you find a way to do this with just the shots looking down and the shots looking up, using just 8 cameras, I would be giddy to see the result.

Breaking Zephyr has been a goal of mine from since I started using it. If you can pull this off, I'd work myself to the bone trying to duplicate it.