Support Website Contact Support Blog

Ray cloud uncalibrated camera

Hi can anyone help?

I’m trying to calibrate uncalibrated images.

Problem A. 1. select the uncalibrated image from LHS list. 2. I click on one of the red X to generate the images below as suggested tie points. 3. Problem the suggested tie points are on a different side of the building no where near the red x.

Problem B. In other images where I believe I have followed the process to create MTP’s the points are indicated by a pink x not an orange x and therefore are not inliers. I need to create 20 inliers to activate camera recalibration.

can anyone suggest why I am having this problem that the software is suggesting tie points on different sides of the building?

Nathan

Hi Nathan,

Could you please attach your quality report to your reply?
(You can attach files using this option:
image )

I would first have a look on your quality report and maybe suggest you some processing options that would calibrate more cameras.
Manual calibration is very painful…

Thanks and looking forward to have a look on your QR.
Best

Hi Marco,

I cant tell if the report attached as an upload. When I tried to drag and drop message read new users cant upload?

Nathan

Hi Nathan,

You are right. This was disabled for new users.
By now you should be able to attach it here.
Sorry about that.
Thanks.
Best

hi Marco,

both of these have uncalibrated cameras. Any help is appreciated.

ferrybridge loader & dolphin_report.pdf (321.6 KB) front left_report.pdf (203.2 KB)

Hi Nathan,

For Ferrybridge, we can see that few cameras are calibrated and the camera optimization is not good.
image

This is due to a poor matching:
image

Which overlap has your flight acquisition? Maybe it is not high enough to achieve good reconstruction. We can see that in the upright corner there no matches founds.

However I can suggest you the following in order to improve the quality of your project:

  • Set the image scale on 1/2
  • Set the Internal Parameters Optimization on “All Prior”

For the other one, we can see that your overlap is good and you have a lot of matches except in one zone where it is poorer:
image

And the camera optimization is very bad:
image

In order to improve the quality I can recommend you:
First, set the matching images pairs on “Aerial Grid or Corridor” and use “All Prior”
If the results are not better, try to reduce the image scale.
I would also, depending on the nature of the surveyed area (if it is flat (usually fields)), use “Use Geometrically Verified Matching”.

Let me know if you get any better results.
Best

hi Marco,

Thank you.

I am using pix4d mapper pro desktop on Mac Pro OSX10.14.3

  1. I cannot see or find any reference to how to select all prior:
    https://support.pix4d.com/hc/en-us/articles/205433155-Menu-Process-Processing-Options-1-Initial-Processing-Matching

  2. Camera optimization: what usually causes this level of error? I have chosen the correct model and focal length.

Nathan

Found the ‘all prior’ selection.

Oh dear the results are way worse by selecting ‘aerial grid’ and ‘all prior’ AG & AP. The image is curved from left to right and sloping back from bottom to top.

Should I have reloaded the file from the start with AG & AP?

Nathan

Hi Nathan,

The All Prior options forces the optimization the internal parameters (principal point, focal length, distortions) to stay close to the initial values.
This option can remedy bad calibration in case that:

  • The terrain is flat and there is no 3D information to facilitate the computation of the optimized camera parameters, which leads to ambiguity (multiple possible solutions).
  • The project is bent after processing.
  • The focal length is optimized far from the initial value, which also indicates ambiguity (example: the cameras are on the level of the model).

But apparently in your project it it did not help.
I am surprised to see that Aerial Grid causes this curvature.

I would suggest you this post as well:

Which file are you talking about?

My recommendation would be here to do several tests on step 1 to see what are the best processing option for you data set. (it is difficult to give a ready to make recipe in photogrammetry. Sorry about that).

Best