Hi there,
I have a use case here that requires the photogrammetry of RED camera footage taken from a helicopter rig. There are 4 RED cameras in total angled in a way that we can later stitch the material of all 4 cameras in nuke to create a dome theater film. The cameras shoot in 8K (36 mega pixel) and have a slight albeit not huge overlap for stitching and blending. I was able to generate JPG sequences from the footage for the photogrammetry in RC and have had varying results. The issue is, the more frames I try to align, the more RC goes wrong. Ideally the camera paths would always be the same with constant offsets between cameras A to D, and as long as the sequence is short enough, that seems to work most of the time. But when I try to align a longer sequence like 1000 - 2000 frames for each camera, then RC produces something where you see the 4 camera paths go off the rails and partly go perpendicular to each other or jumping what would be 500 meters from one frame to the next.
What bothers me the most is. I know the exact position offset and angle of each camera in the rig and so in theory I would only need to align the center camera and somehow tell RC the offset of the other cameras, but I believe there is no way to achieve that right? I can align every camera as a single input and that works, but then I have 4 components that I would need to merge. Could that be a feature worth pursuing?
The way I think of it is the following.
-
You define one group of images as your master camera group
-
You define as many other camera groups as you have in your camera rig and tell RC what the position and angle offset it has to the master camera group.
-
You would of course need matching image numbers in all groups, so that you know that CameraA_image.0001.jpg in your master group was taken at the same time as CameraB_image.0001.jpg in the next group and so on.
-
Then you would align your master camera group and all other cameras would be aligned aligned automatically, or at least would be in the ball park of where they need to be. Then RC could do refinements.
I would really hope for such a workflow to be possible in the future, as unfortunately something like RTK isn’t something that will work with RED cameras or footage. Once I convert the RED files to JPG’s I don’t even have any EXIF data so already have to enter the focal length manually.
Maybe this idea could be picked up. I’d highly appreciate it.
Cheers
Guillaume