I have two questions.
What do the icons circled in red mean in the image below? There is nothing mentioned about these in any video or documentation so i have no clue what they mean.
Why does RC not use all my images. They have a good amount of overlap and Agisoft seems to have no problem using all the images and aligning them. RC seems to be inconsistent on it’s aligning process.
Hi john.martini
Set the ALIGNMENT settings like in this screenshot and try to align the scene. Delete all components except the biggest one, but you need to switch the 2Ds view ( thumbnails ) to 1Ds view so that you can see what is inside the project and so you can see the COMPONENTS.
Alignment_settings_80k.jpg
john.martini wrote:
What do the icons circled in red mean in the image below? There is nothing mentioned about these in any video or documentation so i have no clue what they mean.
Images with the green bullets in their upper-right corner have some common points.
The numbers in the little green rectangles represent number of tie points.
The yellow triangle with an exclamation mark in it means that your camera objective is not in the current software database, which is not a real issue…
Maybe a good idea to add the Milos aligning numbers as default in RC? I also had a lot of aligning problems where, most of the times not always, these settings helped. Agisoft aligned all of them without much problem. But what I got explained, Agisoft is more forgiving at align photos that maybe shouldn’t be in the project when it comes to the quality of the pointcloud generated.
One other thing I had some issues with when it comes to aligning, on some scans the output becomes bend/curved, especially on ground surfaces and flat objects. And on other scans the cameras align in a completely straight line. Read some post here on the forum about selecting the main images “folder” and press grouping by exif could help. On my later scans I have done that and it seems to help. Why isnt that set as standard right away and why wouldn’t you want to group them in the first place?
Hi kallehelgesson
As for the alignment settings, in most cases the default values work flawlessly. The screenshot is a “safe” setting that will work in 99% cases. So changing default values is not necessary.
The main issue that we observe is the quality of the proper capturing process, as Agisoft tends to align almost everything, so you cannot learn where you do it not properly. I can say from my experience from 2013 when I have started testing RC, that it has helped me improve my capture skills SIGNIFICANTLY, as I could learn from it. And the alignment algorithm back then was even weaker for good data than it is now, so it was perfect for mastering my skills. Now I get aligned 99 % of my dataset on the first run.
And almost all issues you described can be pointed to not good capturing scenarios. There are multiple reasons why the “hints” to improve the alignment are not used. Later we will update the algorithms for alignment, but we get feedback from many users that they are happy how it is working right now, because it helps them improve their skills in photogrammetry capturing part of the process.
At least I am personally planning to organize workshops around EU countries this summer so that you can learn how to do it the best way on some real stuff around you.
Hi Milos,
I am new to RC and very impressed of the results so far, but with the current lack of documentation it’s a bit hard to understand the workflows and optimisations.
Thanks for your above explanations but how exactly you deal when the alignment results in multiple components :
I understand from other posts that aligning components is possible but how? and how can we merge them?
In addition, is there a way of optimizing the alignment once its done or it already done by RC and no more can be done ?
Regards
Fred
Frederic LUCAZEAU wrote:
I am new to RC and very impressed of the results so far, but with the current lack of documentation it’s a bit hard to understand the workflows and optimisations.
The Help section built in the program is almost complete, ready to be released soon.
Frederic LUCAZEAU wrote:
I understand from other posts that aligning components is possible but how? and how can we merge them?
You can merge components by providing an additional set of input images, using tie point control points or ground control points and re-aligning the model afterwards.
