Hi,

> William F Pokorny writes:
>> Francois LE COAT wrote:
>>> I have a webcam on my Vaio ... I've moved my head in front of the
>>> camera, extracted the pose of the head with projective transform
>>> parameters, and re-injected those parameters in POV-ray in order to
>>> move a "virtual head" rendered with it. This give the video :
>>>
>>> <http://hebergement.u-psud.fr/lecoat/camera_fixe.mp4>
>>>
>>> Does somebody already seen that kind of demonstration results ?
>>>
>> Neat! I do not recall seeing this sort of camera to model coupling bei
ng
>> accomplished with POV-Ray.
> 
> The processing is only based on computer vision and image processing,
> with one webcam. I obtained the parameters of the movement with image
> registration and an original projective model. I don't think this
> result can be obtained without the POV-Ray scripting capabilities.

A WEB page was made to illustrate the Head Pose experiment...

When we want to determine the movement, there are two simplifying
hypotheses. Either the camera is fixed, and the observed scene is
moving. Or the camera moves, and the scene is static. In the general
case, the camera and the scene are moving, and it is necessary to
segment static and dynamic elements of what is observed. In this case
the camera is fixed, and it observes the person who is located in front
of the computer and moving. The goal of the experiment is to reconstruct
the visible relief, by "monocular depth"...

<https://hebergement.universite-paris-saclay.fr/lecoat/demoweb/profondeur
.html>

That is to say we obtain the depth (inversely proportional to the
disparity) by measuring the optical-flow, with a single camera, tough
this measurement is classically made by binocular vision (for disparity)

Best regards,

-- 

<https://hebergement.universite-paris-saclay.fr/lecoat>

Post a reply to this message