Autocalibration

[rjanvier]

Hello Pierre,

first, thanks for your work on openMVG (and also for Open sourced it !). I have a very simple question regarding the evolution of openMVG: Do you have some plans to implement a kind of autocalibration (like in Apero or in Bundler) in OpenMVG ?

Regards,
Romain

[pmoulon]

Hi Romain,

Not for the moment.
I prefer consider one day adding a six point based solver (For pose and Focal length estimation).

Do you know how is done Autocalibration in Apero ?

Regards,
Pierre

[rjanvier]

thanks Pierre,

I made a deeper look into openMVG source code tonight and now I found my previous question was obviously stupid. In fact Apero, can initialize intrinsic parameters with EXIF data as Bundler and... openMVG. But my error lead me to another question. From the test i've made until now, i'm very impressed by speed and accuracy of openMVG and it's also a clean, maintained and organized code. To unveil my plan, as an ArcheOS ;) maintainer, I'm studying opportunity to replace the good ol' Bundler by openMVG in the PPT pipeline. But Archaeological photogrammetry surveys can make extensive use of wide-angle lenses or super-wide zoom (I'm not talking about fisheyes, I know that's a different kettle of fish). These optics can produce strong distortion... so what about estimation of additional intrinsic parameters in openMVG in the future (ok, sorry if it look like a Santa Claus letter !) ? Or do you recommend to use a software with camera and lens profile to perform distortion correction before ? Or did I miss something again...?

I know it's off-topic and, as Andrew, I'm sorry to report that as an "issue", it's only by pure curiosity... what's you're feeling about replacing Bundler with openMVG in the PPT ?

Regards,
Romain.

[pmoulon]

Hi Romain,

Thanks for the remark and the congratulations.

In order to plan an integration in ArcheOS it missed two main point (one you have cited).

-> The SfM demonstrator support actually only a sequence acquired with one focal lenght. I plan to extend the chain when I will have more time.

-> I prefer that the distortion is corrected before. I have to investigate the quality of the OpenCV calibration engine in order to see if we can do a solution by ourself to propose a calibration of the camera and the lens in order to distort the picture and run the SfM chain safely. (We can also look to see the existing openSource solution that exists)

My feeling about replacing Bundler by openMVG ?
I'm for this kind of change, it could be an excellent opportunity for some student to add functionality.
But yet it's not feasible, the two point cited before have to be managed.

Regards,
Pierre

[rjanvier]

Hi Pierre,

Thanks for this detailed answer ! I found in the Ceres solver documentation that It shouldn't be difficult to add more cameras parameters (at first glance...). Maybe If I have time and necessary skills on my side I can try to do something on the two "problematics" points. But I'm afraid that I couldn't build something strong on this topic, given that I came from Humanities :). For now, if you are interested by an "exportToMicMac" function I think I could handle that...

with all my support regarding your work,
Regards,
Romain