The size of coded markers for UAV acquisition

[ajam13]

Dear all,

We are planing an UAV acquisition at a height approx. 75 m & the GSD = 2 cm. The question is, how big should be the coded marker (printable from PhotoScan) to be automatically recognised and coordinates precisely computed?

It is stated in the Tutorial (Intermediate level):Coded Targets & Scale Bars in Agisoft PhotoScan Pro 1.0.0 that the "central black circle-point on the taken photo is not greater than 30 pix." What would be the minimum requirement (the min no.pixels that the target is recognised)?

There was a discussion in 2012 (Automatic GCP identification), when a right/better marker was searched for .. to be used for mid&far range photogrammetry. Any solutions, we could use now?

Many thanks for any info & advice.

bw, Maja

___________________________________
Maja Bitenc, MSc
University Assistant

Graz University of Technology
Institute of Applied Geosciences
Rechbauerstrasse 12, 8010 Graz, AUSTRIA
Email: bitenc@tugraz.at
Web: www.egam.tugraz.at

[Alexey Pasumansky]

Hello Maja,

Actually, coded targets are not intended to be used for aerial surveys. In the next version of PhotoScan non-coded target automatic detection will be supported for the four-sector circles with white and black sectors.

[bigben]

but in theory...
Print some out at different sizes of a few cm in size and photograph them with the camera you'll be using at different distances. 
Load the images into PS and see which markers get recognised and which ones don't. 
From the pixel dimensions of the smallest recognised marker calculate the angular width of the marker.
Then use this angle to calculate the required width of the marker when photographed from your flight altitude.

[ajam13]

Dear Alexey and bigben,

Thank you for your answers.

Alexey, when do you expect that the automatic target recognition for aerial survey will be ready? Is the curent algorithem for coded marker recognition based on a simple image matching (correlating a photo & the theoretical bit-image) and thus looking for the highest correlation coefficient? Or in other words, is there any documantation how this algorithm works?

Bigben, by saying

Then use this angle to calculate the required width of the marker when photographed from your flight altitude.

you suggest that the PS coded marker size (defined by experimentaly calculated angle) simply increases linearly with the flying altitude? Thus, I could just make the experiment on ground at e.g. 10 or 20 m, having different size targets and then expand the best recognisable target for the planed flying height?

Thanks again!

bw, Maja

[jtuhtan]

Hi,

We use white matte markers which are 50 x 50 cm with a large X in the center, and numbers written in black on two sides of the marker to distinguish 6 from 9, etc.. Most of our flights are from 100m and we have an effective image resolution of 2 cm/px.

Under decent lighting, I have had no problems identifiying the markers from the aerial imagery, but thus far the marker location has been performed manually.

Best,

Jeff

[bigben]

you suggest that the PS coded marker size (defined by experimentaly calculated angle) simply increases linearly with the flying altitude? Thus, I could just make the experiment on ground at e.g. 10 or 20 m, having different size targets and then expand the best recognisable target for the planed flying height?

Not quite.  The minimum size of the markers will a fixed size in pixels, which you need to convert to an angular measurement.  This will depend on the resolution of your sensor and the fov of your lens. Once you determine what that angle is for your camera/lens combination, and simplifying the math to a reasonable approximation..

Width of marker = Tan(Angular diameter of marker) x Height of camera

[ajam13]

Thanks Jeff to share your experiences! Do you maybe have an enample of you "x" target at hand?

I see bigban. Jp, I was thinking about min no. pixels/targed needed to identify target, which is the same as your min angular width. Experiment would for sure be the best way to find it, if only we have time for it

Maybe PhotoScan will implement new targets for aerial surveys and algorithm for their recognition inbetween ?!

lgm

[mihaib]

Hi to all,
in our project we use A4 and A3 size markers!
camera nex6@ 35mm (lens 16-50mm)

[mihaib]

.... and we use a total station to measure the markers, to compare.

[AndyC]

I'm looking forward to the increased functionality for the coded targets; is it possible to get a rough idea when the next PhotoScan version might be released?

[mr map]

... In the next version of PhotoScan non-coded target automatic detection will be supported for the four-sector circles with white and black sectors.

Alexay,

Automatic aerial targets is a great feature 
What will be the minimum recognizable size of the target circle, in pixels?

Tom

[Helsky]

Dear Alexey Pasumansky,

has there bin any update on the topic. I.e.:
- Is there a formula to that allows to calculate the marker size and central point radius in relationship to the ground resolution of a camera
- When will the non-coded target recognition be available?

Thanks a lot and kind regards,
Helge

[Alexey Pasumansky]

Hello Helge,

Two types of non-coded targets are already supported for automatic detection (after finishing of Align Photos stage) in the version 1.2.

[Yoann Courtois]

Hello Alexey and others !

I would like to know more about the minimal size a marker can have on a picture to be automaticly detected with confidence.

I've made many reasearches on PhotoScan manual, help page, forum. But none of them give any number about the minimal size. Only maximal size is sometimes given.

I would like to have informations about non-coded target and also coded target. I'm aware that coded target are not recommended for arial surveys. But I would like to know if a protocol can be created for low flight (precise flight survey).

As well, all of numerical size (maximal size) I found during my researches talk about the size on the central point of the target. Is the only central point which is detected by PhotoScan ? And the rest of the target is only used to code the marker ?

Thanks for you help

P.S. I'm sure I will, at some points, make some detection tests. But I would firstly like to smooth down the topic by having informations about:
What is exactly detected (central point or whole coded target) ? , and
what is the theoretical minimal size which can be detected ? (for exampl 10 pixels of diameter for the central point to be accuratly detected)

[Yoann Courtois]

Hi everyone !

After managing some tests with different cameras (UAV DJI Phantom 4, ActionCam GoPro Hero 4, ...),
I was able to define the minimal size a coded target should have in order to be detected !

Tests parameters :
I have handled my tests with 12 bits target, because they present, according to the user manual, the most ability to be detected.

Result :
In order to be detected, the center point of a PhotoScan 12 bits target have to cover a minimum of 9-10 pixels on the image

Application :
For instance, for a photogrammetric survey with a defined GSD of 1cm, the center point of the target should be around 10cm minimum (10 x GSD). As the global diameter of the target is 3.5 times the diameter of the center point, the whole target should have a minimum size of 35cm. (35 x GSD).

To be printed :
As the center point diameter should be egal or greater than 10 x GSD, the Center point radius should be set at 5 x GSD or more.

For non-normal view :
I didn't handled precise measurements but I can say the angle of view is not the main limitation in automatic detection. For UAV survey with nadiral images for example, the limit of detection will be the effective GSD on the target when the camera will fly away from the target position, and not the angle a view.

Hope it can help