How could I imagine GCP weights. If I set it to for example 1000 does it mean that the aligned point is more tight to the source coordinates or more loose than if I set it to f.e. 1?
It would be better to be able to set weights in form of std to be able to better determine precision of the point.
Grundmann wrote:
How could I imagine GCP weights. If I set it to for example 1000 does it mean that the aligned point is more tight to the source coordinates or more loose that if I set it to f.e. 1?
It would be better to be able to set weights in form of std to be able to better determine precision of the point.
Generally, it is recommended to leave them their default weight value. Number 1 means common strength / importance. The higher the number, the greater the influence of a point on a quality of model alignment.
What do you mean by “std”?
I mean standard deviation or something similar is RMS error. In case I will want to use RC for surveying purposes I will need to determine precision of the gcp coordinates in such a way.
Grundmann wrote:
I mean standard deviation or something similar is RMS error. In case I will want to use RC for surveying purposes I will need to determine precision of the gcp coordinates in such a way.
It is possible to import coordinates of ground controls into RC. You can even test them via using ground test control points.
Hi , it is a good question, what is the weight limit ? I mean, is this logarithmic, or does it have absolute values?
I want to force model segments to match at control point - it often ignores it
Thanks!
Hi Simba,
why don’t you just try it out? Maybe with a smallish project.
I used 1000 just today and it worked - kind off because it does screw up everything else and so it doesn’t really help much.
In my experience there is a problem with alignment if the GCP is not where it’s supposed to be. When I started out I often had errors of several mm (with buildings surveys), but since I know how to do it properly the errors went down by a factor of 10.
I noticed that trying to force RC to do something it doesn’t want to (it’s stubborn like that), won’t really get you what you want. Not too long ago, with my good alignment technique, I had one point that would not work never mind what I did. I almost gave up on it. Until I realized, that RC wants to place the GCP exactly 1 cm lower than the measured coordinates of the point. Then it dawned on me: I sometimes use a tool when I want to measure points on the floor that are far away, which is not possible directly. It’s basically a vertical face with a point 1 cm higher than the ground. So obviously I forgot to adjust the coordinates accoringly and RC was right on the spot! That really convinced me of the incredible accuracy that is possible with RC…
Thanks for your reply
I dont have ground control points, Just tie control points that I want to use to align cameras, but it doesn’t seem to want to do it.
I can see the matching points in the photos, which I make a CP with but it doesnt like to put them together.
Ah, the post was about GCPs - that’s why I assumed you use those too.
It doesn’t change the principle though, since GCPs are just Control Points (which are manual Tie Points) with coordinates.
So what I hear is that you want to “force” images or components together, that won’t do it by themselves.
Sounds like the images aren’t ideal. It would help if you posted some screenshots of the components with the cameras.
thanks gotz
Please find attached and example. I have 4 lines of cameras, but I cant get them back into alignment. On the left they are ok, but the tracks drift off, even if I try to describe the CP
Oh lala! They are supposed to be parallel? 
It’s a bit curious because you seem to have a lot of overlap along the tracks - is there maybe not enough sideways?
Did you try exif grouping yet?
Oh and I just remember that sometimes it is necessary to delete all components in a case like this. RC likes to hold on to previous results as long as they are present in form of a component. But I would save the old components beforehand because after alignment there is no possibility for undoing.
Haha yes!
I noticed also you have spent some interesting time learning and helping people with this SW, as I am going through the forums now to find more. There isnt really enough in the help file and not much else, a few vidies on youtube.
I will try exif grouping. It would be nice to enter the lens data for the camera, its a gopro, I know the K1,K2,K3 values.
I notice that at one end it seems to start is better - I wonder if it is important the order that it reads the files? I read in one place that it definetley doesnt, then someone else found it did.
I need to find a way to force some of the cameras in place, and let it do the rest
Hey Simba,
hehe, so you know why I’m doing it! To learn as I try to help others…
Anyway, it’s hard to judge without seeing some photos.
Have you tried to align smaller batches of images? You can do that by selecting some cameras from your misaligned component and export it as Images List, then import these images as Image Selection. If that works, you can bring them together later. Make sure some images are in both adjoining batches.
I think you can do the K value thing with an XMP file, but I know no details for that. The advantages are marginal since RC needs to calculate the values in the process each time anyway, so you can just optimize the starting point. But I guess that your camera is included in the database by now since it’s fairly popular and RC does it automatically since a while now.
About the order: I don’t think that it matters but you never know. It’s so complex that it might be true in some cases and in others not. I definitely know that there is some randomness to the results of the aligning process. You can see that if you try to align a problematic image set - in some cases it will come up with a different solution almost every time. I am talking about seriously difficult ones. With a good image set, the results will be almost identical every time (I only had a few exceptions).
Exif grouping will help RC in the initial alignment since otherwise it assumes that each image has a different distrotion value. That is not too much of a problem in perfect image sets but can lead to difficulties in problematic ones. With grouping, it creates an average distortion for all images in one group which in a fixed lens will be almost perfect (apart from differences due to focus, aperture etc.).
And finally to quote one of Wishgranters most popular advice:
Take more/better images! If you can…
It’s frustrating to hear but true and very often a lot quicker than manhandling the existing set. A good image set will work on the first alignment, the second at the most. I had to learn that the hard way! But maybe everyone has to go through that stage… 
Thanks Gotz , thats great. You are right , in that there seems to be an element of randomness.
Unfortunately I had to take the photos under v difficult circumstances , so I can only use what I have - but if it happens again I know what to to do!
Question: it is it better to preprocess the photos, eg contrast etc. I saw one tutorial for photoscan , where they increase the contrast for better resolution and scanning. But then I have also been advised to leave them alone since (another) scanning software will only use what is there
. Mine are in a kind of RAW format (GOPRO - ‘protune’) - which means they ‘look’ flatter , but in fact contain more information (like a RAW photo looks worse, until you bring it into photoshop and edit, then it would look better than JPG)
All the best
Hi Simba,
I only got the message now that you replied, so sorry for the delay.
I know what you mean and I have read similar condratictory advice. It depends on so many factors, that you will pobably have to try it out for yourself. What I do also depends on the object. I always correct abberation, if needed vignetting and sometimes brigthen dark colours very carefully. With contrast I don’t mess.