Hi, today I took and imported a continuous shot into Reality Capture, however the program split it up into components, and I do not understand why as there is no obvious difference between the two frames it split at, here are a couple examples
For the most part it has detected a sequence but I don’t understand why its split it at this point, feels pretty random to me…
Also want to ask why do things like this happen in a component?
Again no obvious differences with the offending cameras
And also why does this happen, again see no obvious defect in the frames… (can’t include pics here of this as then that would be too many, but this is as the road i was filming went uphill)
These feel like beginner problems, of course it might be my footage but i am struggling to see how other bits from the same shot work fine, and i can’t see any obvious differences from that and the ones that arent working… If its not the footage is there anything that will help the program understand its a continuous shot (I wish control points were properly in the pre-alignment part)
What you can also try is to make more photos from a different angle. So take some additional photos from the other side of the street or less close to the wall, which is visible in your photos. Get a overlapping for each photo of at least 30%
Also try to get sharp images, with a mid-range tonemapping. So not to bright and not to dark. everything what´s to bright is lost. Take RAW - images if possible. It has more image information and you can make small corrections in a photo editing tool if needed. But a unsharp image, will be still unsharp.
Thanks everyone for the responses. I have tried adding control points before alignment, and this is weird… The place where I put the control points, and I literally only put 2 for about 20 frames, now aligns perfectly as if there was no issue, but now it has split elsewhere at a point that was previously aligning fine… Why could this be… Its as if the program thinks it has to split it up and the control points have told the program not at that particular point…
This must be a settings problem, these are my settings
Turns out the problem was i did not set the calibration group, but still it is separating at a couple of points when a frame gets blurry or there is an abrupt change of direction, is control points the only way to fix that?
I can say things have got a lot better with the reconstruction, however there is one small hiccup, why does this happen even though the cameras are aligned and they are calibrated to the same size? And how do I fix it as its not obvious what cameras its linked to?
As this is not ideal dataset for photogrammetry (it is one straight line) I would use some CPs in those problematic parts. You can use lasso tool to select tie point in this area and then show the cameras related to those points. Also, georeferencing could help there.
Control points did work but I had to realign the whole thing and redo the control points i needed to put in to stitch the separated components together, is there a way I can simply just realign the selected part?
It is better to realign a whole part to new control points, as you want to have one model. There are relation between the images, which are improved by used control points.