POV-Ray : Newsgroups : povray.advanced-users : Skewness (long) Server Time23 Jun 2024 06:53:13 EDT (-0400)
 Skewness (long) (Message 11 to 11 of 11)
 From: Bald Eagle Subject: Re: Skewness (long) Date: 26 Nov 2018 21:20:01 Message:
```
{
"@context": "https://schema.org",
"@type": "DiscussionForumPosting",
"@id": "#web.5bfca9aa1f36de80765e06870%40news.povray.org",
"dateCreated": "2018-11-27T02:20:01+00:00",
"datePublished": "2018-11-27T02:20:01+00:00",
"author": {
"@type": "Person",
"name": "Bald Eagle"
}
}
Francois LE COAT <lec### [at] atariorg> wrote:

> I made a WEB page to explain how to approximate the 2D transformations
> of the images, in order to model 3D position of the camera in space :
>
> <http://hebergement.u-psud.fr/lecoat/demoweb/camera_localization.html>

These are very impressive results indeed!

I myself cannot see the "how".

I must admit I would be extremely challenged to derive the following 8 matrices
given only those 2 photos.

Projective transformation:

Tx (translation) = -2.0 pixels
Ty (translation) = 5.9 pixels
Tz (translation) = 1.6 pixels

Correlation = 82.25%

Are those matrices calculated and applied in that order?
It would be beautiful to see a short step-wise derivation of how the distorted
image is corrected to match the reference image, and how the matrices are
determined from comparing both images.

Do you use some sort of image processing to obtain reference markers?
How do you automate the alignment and image-overlap correlation?
How many software packages do you use, and how long does it take to match the
two photos for each frame of the animation?

> All along the video sequence, the 3D localization of the camera is
> estimated, thanks to the image contents of the sequence, and the
> position is represented with the POV-Ray transformations. 8 transforms
> are used : <Tx,Ty,Tz> translations in pixels, <Rx,Ry,Rz> pitch, yaw and
> roll rotations in degrees, and <Sx,Sy> skew (or shear) angles in degrees

I very much like the additional camera object - it gives some better intuitive
idea of what is taking place.   Very impressive that you transform the image and
then back-calculate the attributes of the camera!  :O

With the results that you have, is it possible to calculate approximate
coordinates for the edges of the buildings, windows, etc?

> I'm happy that it works, because the movement of camera is very large =

I'd be ecstatic with your results!

> Thanks again for your help.

Ha - it doesn't seem like you needed much!  ;)

Thank you so much for returning to the forum and sharing the progress you've
made on your work!  It is very satisfying to see what you've accomplished and
see the beautiful animation that you've put together.

Truly brilliant work.  You should be proud!   :)
```