Forum

Author Topic: Photogrammetry with targets  (Read 3890 times)

Cyberworld

  • Newbie
  • *
  • Posts: 38
    • View Profile
Photogrammetry with targets
« on: November 18, 2020, 12:53:08 AM »
Probably a question that common sense already answers, but here it goes anyway: I'm working on object photogrammetry and I have targets of known size (5mm) around my subject in order to be able to scale the object during processing. Question is: What are the methodological/practical requirements to use these targets in Agisoft Metashape to create scales out of them? More specifically:
- If I mask the images to exclude background, can the targets be in the masked region or do they need to be in the region that gets processed?
- If the targets need to be in the region that gets processed, can I later delete them in dense cloud or mesh stage or if I do then scales won't work?

Thank you for reading the question and any help provided. :D

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Photogrammetry with targets
« Reply #1 on: November 24, 2020, 12:58:02 PM »
Hallo Cyberworld,

perhaps it helps to have a look here: https://agisoft.freshdesk.com/support/solutions/articles/31000148855



and here https://www.agisoft.com/pdf/tips_and_tricks/CHI_Calibrated_Scale_Bar_Placement_and_Processing.pdf


To your questions:  You can later cut out the targets. For that it is of course better that the targets are next to and not on your object.


Best regards

Karsten
« Last Edit: November 24, 2020, 04:30:38 PM by Kiesel »

Cyberworld

  • Newbie
  • *
  • Posts: 38
    • View Profile
Re: Photogrammetry with targets
« Reply #2 on: November 25, 2020, 02:28:10 PM »
Thank you Kiesel for the links. Unfortunately they are not what I was seeking, because they are dealing with coded targets, which are not used in my case. However, I did some experiments of my own and for the benefit of forum members these are the findings:
- Either with or without the targets masked, after you run initial picture alignment Metashape is capable of detecting non-coded markers residing within the masked areas of the photos used. The software identifies and marks them correctly in space.
- In my case I use four groups of black-on-white targets, each group comprising of four equidistantly placed targets (4x0.5cm black circular targets on white backdrop, equidistantly placed at 1.5cm each from the centre of the other in a 2x2 array), hence there is a total of 16 targets. The four groups are placed at the four edges of each object captured. Since the software recognises the centres of each marker and the distance between any two markers within each group is a constant, placing scale bars and correctly scaling the object becomes very easy either at the start of the procedure or at the end, as it appears that scaling does not affect calculations during point cloud production.
- Only caveat or point to be careful about is that you have to make sure the targets don't move around while shooting, as movement can increase error in the scales or make it impossible for the software to recognise targets altogether.

Hope the above are helpful.

All best,
Cyberworld

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Photogrammetry with targets
« Reply #3 on: November 25, 2020, 04:14:05 PM »
Hallo Cyberworld.

It might be that I misunderstand you. Do you really scale your object with the help of the 1,5 cm distance between your dot markers in the 2x2 group?
If yes, that is not a good idea to go from small to big, it is allways better to go from big to small for scaling for better over all accuracy. So in your case from one group to another.

Best regards,

Karsten

Cyberworld

  • Newbie
  • *
  • Posts: 38
    • View Profile
Re: Photogrammetry with targets
« Reply #4 on: December 23, 2020, 12:45:13 AM »
Do you really scale your object with the help of the 1,5 cm distance between your dot markers in the 2x2 group?
If yes, that is not a good idea to go from small to big, it is allways better to go from big to small for scaling for better over all accuracy. So in your case from one group to another.

I indeed scale using the dot markers, but keep in mind that the objects are never too big (max 50cm, most around 20-30cm). I've also calibrated and properly measured the distances in mm between the 2x2 group centres to 3 decimals for increased accuracy.

Why is it an issue to go from small to big?

Cyberworld

  • Newbie
  • *
  • Posts: 38
    • View Profile
Re: Photogrammetry with targets
« Reply #5 on: December 23, 2020, 07:29:55 PM »
So in your case from one group to another.

This is also feasible, since the photos are at 50MP resolution, with the targets not moving the entire shoot and being spread at the edges of each object digitised, hence it is possible to also extrapolate the distances between groups (centres of outer circles in each group that is) based on the known distances within each group.

Thank you for your help on this. It is much appreciated! :)

Kiesel

  • Sr. Member
  • ****
  • Posts: 332
    • View Profile
Re: Photogrammetry with targets
« Reply #6 on: January 05, 2021, 10:40:05 AM »
Quote
Why is it an issue to go from small to big?

Because you scale the error of your measured scale bare as well.

For example in your case the error of your measured 15mm has to be multiplied by a factor of 20x when you have an object size of 300mm and with 40x for an object size of 600mm. When your accuracy of your measured scale is +-0.5mm you get an scale error of +-10mm for your 300mm object and +-20mm for your 600mm object.
In the opposite case from big to small you have a scale factor smaller than 1x, so your object scale is better than your accuracy of your measured scale bar.
But all depends how accurate you can measure something (big or small).

Btw. another advantage of coded targets is that those can be measured with subpixel accuracy up to1/20 of a pixel.


All the best

Kiesel
« Last Edit: January 05, 2021, 01:56:13 PM by Kiesel »

Cyberworld

  • Newbie
  • *
  • Posts: 38
    • View Profile
Re: Photogrammetry with targets
« Reply #7 on: January 27, 2021, 06:17:34 AM »
Quote
When your accuracy of your measured scale is +-0.5mm you get an scale error of +-10mm for your 300mm object and +-20mm for your 600mm object.

Thank you for the clarification. The measured accuracy I have is +-0.005mm. But in any case, I've managed to use computer vision on photos to also calculate distance between targets, therefore I also have large scales, which I hope will also be useful.