I would like to request a new tool named “Alignment Tool”.
The tool could be used to guess the best alignment settings.
You could define 40 images in your image set that are surely overlapping and which give you trouble because they are split into 2 components while they should be only 1.
The Alignment Tool would run several alignment tests on its own, analyze the results (=number of components and numbers of images used) and display the best tested alignment parameters.
I think this is what I’m doing right now. but I think a tool can do this faster and better than I can.
Thank you for your consideration
Something to read!:
https://support.capturingreality.com/hc/en-us/community/posts/115003012052-Optimising-Alignment
https://support.capturingreality.com/hc/en-us/community/posts/115000790031-How-to-optimize-Reconstruction (really more about Alignment). The guys are suggesting that as well as a one-shot set of Alignment Settings that’s optimum for any given photo set, it’s also a multi-shot process where a) RC ‘trains’ itself from one run to the next, b) the model can be built up a bit at a time, progressively adding more images on a high quality base and c) settings can be progressively modified over several runs.
Hi alice,
in theory, I think it’s a good idea. The problem though is that RC doesn’t know what’s best for you. It cannot judge the content of the images and wheter they are correctly aligned according to that. This is something, that only humans can do (at least for now still). On the other hand, it would probably also be already an improvement , if you had one or two handful of alignments to browse through. I guess you could customize that with CLI. What you have to keep in mind though is that with difficult sets, RC will come up with different results every time even with the same settings. So really what it means when components are too fractured is that your images aren’T ideal. I know how frustrating it is to hear that because it was the same learning process for me. Just a few weeks back I wasted a week on such an image set.! 
Hmmmm. isn’t the point cloud density a good estimation for quality?
Hmm, how did that get in here? 
Not really, a dense point cloud just means that there are more Tie Points, not necessarily that they are good…