Camera station questions

[bigben]

Moving closer to looking at a pole mounted rig. I need to understand a couple of the finer points of camera stations in PS to see how they might fit in with the direction my shooting strategy is taking.

1. Is a small parallax error OK, or does it assume that there is no parallax change between cameras?
2. Are there any specific advantages to processing time and/or accuracy from defining images as a camera station or just leaving the in as separate cameras?

... and any other info/tips thankfully considered...

[bigben]

I've been doing some tests using my panorama gear (nodal ninja, full frame fisheye) and have pretty much answered 2. Big advantages to accuracy. Putting this into practise with a multi camera rig though, it would be nice to know if there is some optimisation of camera positions performed to allow for small changes in parallax or whether it requires no parallax changes within the group of images?

[Marcel]

What do you mean with Parallax changes? Parallax as in the 'no parallax point' for panoramas?

I was using a 2 camera setup (2x Sony A7R) and I was running into a lot of trouble with shutter shock. The shutters never trigger exactly at the same time, so in a lot of photos there was subtle motion blur because of the shutter shock from the other camera. The effect was very subtle, but enough to cause noise in the scans.

So when using multiple cameras I would certainly use a camera that supports EFC to minimize shutter shock, and a very sturdy pole.

[bigben]

Yes. With a single camera and a pano setup the camera is rotated around the no parallax point (I'm old school and still call it the nodal point) of the lens. That's the same in principle as a  camera station group.  If you have a multi camera rig, the cameras are in different positions but very close to each other. For panorama stitching that's not so much of a problem because slight distortions can be introduced to compensate for this, but in photogrammetry that's not practical.  If Photoscan allows for a slight tweaking of camera position within a camera station group then this may be OK, but at some point that will fail and they will have to be treated as separate cameras... what that point is I don't know yet.

Will keep an eye out for shutter shock. Thanks for the tip.

[bigben]

Dense point cloud is still processing but I am now officially excited... again

~200 camera stations, mixture of 35mm lens or 15mm fisheye, nicely aligned with much better geometry than anything I'd done with a single camera on a scene this large.  Definitely worth the effort.

I also reprocessed some images from an earlier pole mounted camera. Back to back circular fisheye images. Camera was orientated perpendicular to the camera path and Photoscan didn't like it even though the images eventually overlapped as I went around a corner. Grouping the pairs as camera stations worked nicely. Didn't processor far enough to see if the camera offset of ~10cm had any impact but I suspect it may be negligible (camera 3.5m above ground, ~10deg masked from bottom of the images to remove me)

Using the pano rig with the 35mm lens was nice as all of the rotations were fixed so I didn't have to look through the camera

[bigben]

Low quality point cloud.... too impatient 

[mrb]

I did some experiments with camera stations vs. traditional "walk-around-and-photograph" capture this past week, using a fisheye lens.  In a medium sized room, I did 3 camera stations, taking 60 photographs total (chunk A).  I then took 55 photographs as I typically do (chunk B).  I processed both as separate chunks to compare the isolated results, and then combined all the photos into a 3rd chunk and processed that (chunk C). 

The camera station only (chunk A) was miserable compared to the traditional chunk (B), and the chunk of the two combined photosets seemed to be getting all its detail from the traditional shots and was almost identical to the separate chunk B.

Would I have needed to do more than 3 camera stations to see any noticeable difference?

In short, I'm wondering: what's the advantage to doing camera stations at all?

[bigben]

Kind of asked an answered really.  3 camera stations = 3 cameras... and you're wondering why it's not as good as 55 cameras.   

If it's a fisheye lens why are you shooting 20 images per station? Full frame fisheye will get you everything with 5 shots (4 @ 90°, tilted -15° and 1 shot pointing straight up) You'll get a hole around your tripod that will have to be filled with adjacent cameras

The camera positions of all of the cameras are calculated as a single point, and since they are being calculated from a wide range of angles, the calculation of the camera position tends to optimise more accurately than individual cameras, this in turn help to optimise distortion parameters (not much benefit if you're using calibrated values) and in my limited experience so far can produce cleaner point clouds.  If you have a low resolution camera you can also use a longer lens and multiple images to increase the effective resolution.

[mrb]

Kind of asked an answered really.  3 camera stations = 3 cameras... and you're wondering why it's not as good as 55 cameras.   

If it's a fisheye lens why are you shooting 20 images per station? Full frame fisheye will get you everything with 5 shots (4 @ 90°, tilted -15° and 1 shot pointing straight up) You'll get a hole around your tripod that will have to be filled with adjacent cameras

The camera positions of all of the cameras are calculated as a single point, and since they are being calculated from a wide range of angles, the calculation of the camera position tends to optimise more accurately than individual cameras, this in turn help to optimise distortion parameters (not much benefit if you're using calibrated values) and in my limited experience so far can produce cleaner point clouds.  If you have a low resolution camera you can also use a longer lens and multiple images to increase the effective resolution.

Gotcha. I figured that 3 stations was the equivalent to "3 cameras", just wanted confirmation.  As to why I was shooting 20 per station - because this is the first time I'd ever tried it, and clearly didn't know better. Now I do. Thanks!

I'm using calibrated lenses for all my handheld work, and I was hoping that the camera station method would work better for doing interiors, but I don't think it will.

[bigben]

It depends on the interior. Nothing will compensate for blank walls.  With lots of furniture you need more cameras to get around all of the furniture, but there is also a closer working distance. Back to back circular fisheye images may be practical in this situation, but the benefit may be limited. I haven't tried this yet and if I did try it I'd also supplement it with some additional shots with the same lens for the really tight spots.

The other thing to remember is that you don't need a lot of overlap between images within the camera group... it will still work even if the images don't overlap provided there is overlap with images from other camera stations (not that I'd recommend doing that).

[ThomasVD]

To BigBen and Marcel - This is perhaps a bit off-topic but I was wondering whether in your experience use of camera stations improves alignment times? For instance I suppose theoretically in a camera station with 5 cameras, as soon as 1 camera is aligned, PhotoScan should be able to instantly determine the position of the 4 other cameras.

What has your experience been?

[Marcel]

My experiments with a dual camera setup was before the Camera Stations was introduced, so I can't tell.