Optimize camera alignment

[guigus60]

Hi,

I'm learning a bit more about the software and the process.
I've been flying above a pit with approx. 600 pictures from 120m above the ground, all my image are looking down.
Why do we need to do a camera alignment? and how to decide which parameter to tick?
When I tick Fit F, Fit Cx,Cy with the same image every time I have a Cx,Cy different. From what I understand the Cx,Cy are the principal point which measure the offset of the camera centre, right? So this parameter should be all the time the same with the same image?

Regards,

[stihl]

You need to align your cameras in order for Photoscan to estimate their camera location. Whether that's in a coordinate system or an arbitrary local system. Photoscan uses an algorithm to detect tie-points in your images and then matches these together. Once PS completes that it knows the estimated camera location relative to eachother. With this information it uses a different algorithm to generate a Dense Cloud.

I believe that in the latest version of Photoscan it chooses for you which parameters to tick.
You are correct about the Cx and Cy parameters. All these parameters are calculated by Photoscan by refining it's initial alignment. This is done so by the button "Optimize Cameras".
In the 'Tools > Camera Calibration' window and then in the 'Adjusted' tab you'll find the calculated parameters.
The values you'll find in that tab are the ones that Photoscan uses and applies for *all* the camera's in that particular camera group. These parameters are further refined every time you optimize the camera's.
These parameters will also change (for the good or for the worse) if you adjust the marker/camera accuracy in the reference panel. Make sure to test a lot to see what works for your projects.

[prashant]

Is there a place where the optimize camera parameters are explained? It is not very intuitive what they mean e.g. cx cy, fitx, etc. Thanks in advance.

[stihl]

fx, fy
Focal length in x- and y-dimensions measured in pixels.
cx, cy
Principal point coordinates, i.e. coordinates of lens optical axis interception with sensor plane.
skew
Skew transformation coefficient.
k1, k2, k3, k4
Radial distortion coefficients.
p1, p2, p3, p4
Tangential distortion coefficients.

Source: http://www.agisoft.com/pdf/photoscan-pro_1_2_en.pdf

[xyz_rgb]

Dear Stihl
I'm just new in aerial photography
Need advises about Radial distortion coefficients & Tangential distortion coefficients.
What the purposes k1? k2? k3? k4 & p1? p2? p3? p4?
I also confused for better GCP input before optimize cameras use projection WGS '84 or UTM Zone?

Thanks Stihl for quick advises.


Obehe.

[Kiesel]

For distorsion coefficients you find some explanations here:
https://en.wikipedia.org/wiki/Distortion_(optics)

and how it is used in PhotoScan in the user manual "Appendix C. Camera models".

Karsten

[SJuriaans]

Hello there.
Please, forgive me for my bad english, but i would like to know if this process (optimize and alignment) is the same for a RTK drone.
I hope you understand.
Thanks.

[Alexey Pasumansky]

Hello SJuriaans,

For RTK coordinates that are considered to be very accurate it is suggested to adjust "camera accuracy (m)" parameter in the Reference pane settings dialog according to the measurement device precision and then optimize the alignment.

[SJuriaans]

Thank you Alexey!