average tie point multiplicity

[mauroB]

Dear all,
I was wondering how the average tie points multiplicity statistic (chunk info) is calculated.
Is it based on the valid correspondences (i.e. those remaining after outliers detection) or on the total found ones?.
I noticed a discrepancy between the reported statistic and the computed (by mysellf) average number of valid projections for each tie point.
Regards,
MB

[Paulo]

Yes MauroB, you are right,

the tie pt multiplicity is calculated from original projections/tie-points. After you do a gradual selection eliminating some tie pts than this average will change....

[mauroB]

Dear Paul,
Many thanks for your reply.
I noticed this discrepancy while working on a script aimed to evaluate the intersection angle between corresponding image rays (i.e. projections) for each point of the BA-derived sparse cloud.
Since you are a "guru" of Agisoft Metashape, could you provide me with a feedback on developed scripts?
The first attached script is for extracting the projections of each tie point within agisoft (along with data on tie points and cameras coordinates in the model, geocentric and world coordinates systems and re-projection errors), whereas the second one is for analysing in a generic python-based IDE the intersection angles (and it is based on the output of the first code).
Many thanks in advance.
Regards,
MB

[Paulo]

Ciao Mauro,

mio italiano non é perfecto, parlo piu francese o spagnolo... ma é un piacere aiutarlo! Su script projections sembra molto bene... ora lanzo il script angles para vedere suo resultado....

Non sono un `guru` sino un canadese vivendo en Messico, malato de la fotogrametria en general y de Metashape en particular,

P.S. ho avviato il script angles en la console de Metashape con un bloque con una nuvola de 121 982 pontos y 446 102 proiezione... Tarda molto. É normale?

[mauroB]

Dear Paulo,
First of all, thank you very much.
Honestly, I don't use the metashape console for doing data anaysis and computation...
Usually I wotk within a pyhton-based IDE (i.e., spyder)...
The script for the computation of basic stats on the intersection angles might take a while since, in its actual basic configuration, it is not computational efficient (there are three nested for loops). However, it should end (at worst) in 10-15' (I think). Attached you can find a new version of the script and a screen capture of its basic output in a generic IDE.

[Paulo]

H:\DCIM\100RICOHery interesting Mauro, Grazie per la nova versione...

[mauroB]

Dear Paulo,
What do you think about the implemented algorithm for angle computation?.
I performed some cross-check within CloudCompare and I found a general agreement.

Overall, I had some doubts relating to the different reference systems used within Metashape.
Basically, it uses three different reference systems (i.e. the chunk or "model or block" reference system, the geocentric reference system and the "world" reference system photogrammetrically speaking), which are basically related through a similarity transform (i.e., the chunk.transform and chunk.crs.project functions in the Metashape's python API, respectively). Therefore, considered this geometric shape-invariant link, it is my opinion that the computed intersections angles should be the same regardless the considered reference systems (as it is, based on the script outputs).

[Paulo]

Yes Mauro,

I did not look in detail at your angle computation algorithm but it seems OK. I think that the angle computation will be the same in internal CS and geocentric CS which are both cartesian. However the angle computation in projected CS (chunk.crs) would be different as the projected CS is not cartesian (XY are scaled by projection scale factor while Z is not...)