size for coded markes

[andreas_IWB_TUG]

I am doing my master thesis with photoscan in an hydraulic lab.
Now i am preparing my scene with markers and I need more details to print them out.

e.g. Marker type 12bit,
Centerpoint radius = ?,
Targets per page = 6

 I've read the the tutorial PS_1.1_Tutorial (IL) - Coded Targes and Scale Bars.pdf...
"Note that the size of the coded targets should not be too big or too small compared to the
object or scene. It is recommended that the size of central black circle-point on the taken
photo is not greater than 30 pix."

Some details from my side:
Camera resolution 6000x4000 pix
Lens with fixed focal length 24 mm
Distance capturing photos is 4.5 m, --> 0.73 mm/pix

I hope you can help me soon!

Many thanks,
Andreas

[Kiesel]

Hello Andreas,

you have done the math by yourselve and only missed the last step 

0,73 mm/pix * 30 pix = 21,9 mm

So a little bit less than that.

Karsten

[Dave Martin]

and Andreas, to get fine control and markers of a specific size, as has been recommended on this forum recently, just print them from PhotoScan to PDF and you can then use your Acrobat reader or similar to vary the size when you print to paper.

Dave

[andreas_IWB_TUG]

Many thanks for your quick reply!

@ Karsten, size:
but how is the size of the central black circle-point defined? Is it diameter, radius or area?
in the tutorial i only can find this upper limit of 30pix or in my case 21.9mm ... when is it too small?

I'm looking for a range of this radius related to ability being detected, accuracy, picture size, image resolution, ...

@Dave, printing:
thanks, i already found this topic. From experience i can tell you that pdf has an error of 1-2% you can't print HQ. You can see the error after printing. For HQ-printing copy shops are plotting directly using the driver of the plotter. I opened the pdf in photoshop and printed directly out of it.

Andreas

[Kiesel]

Andreas, I think it is diameter.

The procedure behind is finding the centrum of the circle/ ellipse in subpixel accuracy and it seems only to work until 30 pix in Photoscan. I guess best range is from 10 to 30 pix, because the code of the marker has also to be detected.

To look what is the math behind you can see for example here:
https://books.google.de/books?id=G8a5BQAAQBAJ&pg=PA221&lpg=PA221&dq=detection+circular+coded+marker+in+photogrammetry&source=bl&ots=I-BDhrIzMe&sig=TyV-WBi1dR_hQ0qPwqQ3RMvVcE0&hl=de&sa=X&ved=0ahUKEwiM4vbamr_JAhVEdQ8KHcIFBMMQ6AEIQTAD#v=onepage&q=detection%20circular%20coded%20marker%20in%20photogrammetry&f=false

Just guessing: If the markers are to big, it perhaps could help to downscale the images for marker detection. I'm at the standard edition here so can't test coded markers detection.

Karsten

[andreas_IWB_TUG]

Karsten,

danke für den Buchtipp!

[Kiesel]

You are welcome.

From experience i can tell you that pdf has an error of 1-2% you can't print HQ. You can see the error after printing.

You don't have to be very precise here because the absolut size is not so important, it differs in your scene anyway (different distances from the targets to your camera).
With markers you can reach at least 1/10th of pixel size, so in your case 0,07 mm. This should be enough ore not? How do you measure the absolut positions in meters of your targets with a comparable accuracy or you don't need to? When 1 mm accuracy is enough you don't have to think about accuracy of your markers.

A good measurement configuration, camera calibration, camera, lens... is also important. Some very good informations could you find in industrial photogrammetry systems like V-Stars, see:
http://www.geodetic.com/v-stars/what-is-photogrammetry.aspx

Karsten

[andreas_IWB_TUG]

absolute position will be measured with a total station and we can expect an accuracy brief below 1mm. an higher accuracy is not possible for me (too expensive).

i'd like to have an accuracy of 0.5mm in vertical direction, that means approx. an acc. in horizontal direction 0.5/4 = ~0.1mm.
If you say 0.07 is doable, this would be great 

thanks a lot!

[Kiesel]

absolute position will be measured with a total station and we can expect an accuracy brief below 1mm. an higher accuracy is not possible for me (too expensive).

How do you reach this accuracy with a total station normally they are not that accurate, by intersection?
To improve accuracy of measurement with a total station to 0,1 mm (if you really need that much) one can also use a scale bar measurement (Basislattenmessung)  below 10 m distance space, not that expensive.

If you say 0.07 is doable, this would be great 

To be clear this is how accurate Photoscan can measure the centre of the marker in one image, this is not the accuracy of the 3D point which is calculated by 'photogrammetric' intersection and depends on the intersection angle and a lot of photogrammetric factors.

Karsten