Messages

1

General / Re: Matches listed in console but not in "View Matches"

« on: February 26, 2021, 12:34:16 PM »

Hello Alexey,

That's it!
Some images were masked completely, because they only showed the calibration board from the side. I ran it again using "Apply Masks to Key Points" and it worked.

Thank you

2

General / Re: Matches listed in console but not in "View Matches"

« on: February 25, 2021, 03:21:59 PM »

Hi Alexey,

Sure, here is the part of the log file related to this alignment procedure.

3

General / Matches listed in console but not in "View Matches"

« on: February 25, 2021, 12:08:51 PM »

Hi,

I can not get the Photo Alignment to work, because Metashape does not find any matches between the images. The weird thing is that the console (or the log file) says that many hundreds of thousands of matches are found, however, the alignment fails and when I click on "View Matches" I see that most images have no matches at all. Just a few images have up to 10 matches and they are all marked as invalid. I knew the issue of having thousands of matches that are marked as "Invalid" in  the "View Matches" window, but in this case I don't even have invalid matches. I can't explain at which steps I lost all the matches that are mentioned in the console. I attached a screenshot to clarify my situation.

I am using 28 sets of Photos from 4 DSLR cameras that I use as a Multi-Camera-System (i.e. I have 4 sensors and 28 cameras). The images are from a calibration board that I believe to be suitable for matching in Metashape, as it contains a lot of texture.

Thanks in advance for any help! 

4

General / Re: Validate invalid matches

« on: January 13, 2021, 04:20:42 PM »

I just noticed something that seems to contradict the theory, that a match can only become valid if the key point is found in at least a third image.

See the attached screenshot, where I used "Gradual Selection" to filter the Tie Points by "Image Count". If I filter for 2 images, I have a bunch of Tie points selected. But shouldn't the number of tie points with just 2 projections be zero, if a match can only become valid if it's found in at least 3 images?

5

General / Re: Validate invalid matches

« on: January 13, 2021, 11:41:01 AM »

Thanks to the explanation by James I found a dirty workaround, that gets me more valid matches.

A set of key points can only be 'used' to align an image if they are matched in two other aligned images.

I just duplicated the whole chunk, renamed some of the images and moved them back to the original chunk. So I now had some images twice in my original chunk. When I did the "Align Photos" step, I ended up with more matches validated.

This is my humble understanding of what is going on:
I have two images A and B and I copy image B and rename it to B'. In the alignment step, all(?) matches between image A and B can also be found between images A and B'. Therefore we pass this margin that James was talking about. Of course this would come at the price of having more false matches. But that's better than no matches at all, I guess.

Not sure if this is really what is happening, but  maybe it helps someone. What do you think?

6

General / Re: Valid matches marked as invalid

« on: January 12, 2021, 08:01:32 PM »

Hey James thank you for your reply.
The need for a keypoint to be found in at least two other aligned images is a very helpful information. Somehow I missed that before. Thanks!

7

General / Re: Valid matches marked as invalid

« on: January 12, 2021, 04:19:31 PM »

This still remains a recurring issue for me. Could anybody help us out with this, please?

8

General / Re: Reprojection Error: Gradual Selection vs. Chunk Info

« on: August 20, 2020, 10:04:26 AM »

Aah, that makes a lot of sense now   Thank you guys!

9

General / Reprojection Error: Gradual Selection vs. Chunk Info

« on: August 19, 2020, 04:27:39 PM »

Sorry to dig up this old question again. I thought I had understood the reprojection error, but when trying to remove tie points with bad reprojection errors I saw that the RMS reprojection error from the chunk info was very different from the biggest reprojection error shown under gradual selection.

I understand that the reprojection error is the distance in pixels of the reprojected tie point to the respective key point on an image.

As you can see in the attached figure, the RMS Reprojection error in the chunk info is 2.46 pixels. But when I go to gradual selection and choose "Reprojection error", the largest value where any tie points are selected is "Level 0.84". I would have expected the gradual selection value to be higher.

My best guess for why this is, would be that the Reprojection Error "Level" in the Gradual selection dialogue doesnt refer to the original image resolution but a downscaled version of the image. Or maybe the two values are calculated over a different range of tie points?
Could you please explain why these values are so different?

10

Camera Calibration / Principal point coordinates always 0 in Inpho.prj file

« on: April 02, 2019, 12:35:33 PM »

Hello,

I have recently spent a lot of time trying to understand the way the interior camera calibration works in Photoscan.
I export the cameras to an Inpho.prj file and I noticed that the values displayed under "Tools"->"Camera Calibration..." (both the adjusted and initial values) vary greatly from the ones shown in the Inpho.prj file.
And I noticed that no matter what values are displayed for cx and cy under "Camera Calibration...", the "$PRINCIPAL_POINT_PPA" parameter in the inpho file is always zero. I tried manually inserting values with the "precalibrated" option and the "fix calibration" option checked but the principal point in the inpho file was still at zero.

Is it in the nature of the Inpho.prj format to ignore the principal point or am I making a mistake?

Any help will be much appreciated!

11

General / Re: Questions about View matches in PhotoScan

« on: February 05, 2019, 03:07:54 PM »

Allow me to dig up this thread again to get some further clarification of the question by Rosy:

Which measure does Photoscan use to judge whether the matches are correct?

I use Photoscan to calculate accurate camera positions using a calibration target with a nice pattern and known geometry. During the photo alignment Photoscan finds around 20,000 matches between image pairs, but marks all of them as invalid. This made me wonder about two questions:

1)  Based on which criteria does Photoscan decide whether a match is valid or invalid? (original question from this thread)

2) What would be the characteristics of an ideal surface pattern, that would lead to a maximum number of valid matches?

Cheers,
Felix