|
|
"melo" <mel### [at] coxnet> wrote in message
news:web.47d9f8f03d86ff96314b3d800@news.povray.org...
> It appears you might have a different/local version of POVPerson.
>
> When I tried your new simple syntax recommendations i.e
>
> #include "povperson.inc"
> ppFigure()
>
> I got a Parse Error saying : Expected 5 parameters but only 0 found.
Oh. Well looking on the bright side, that means that at least the
documentation and the code on the website are more consistent than it
appeared for a moment. :-)
>
> Now a totally unrelated question:
>
> Is it possible to specify movement constraints in POV_RAY.
It's possible to specify constraints on joint rotations to keep a joint
within realistic limits.
> I had managed to
> calculate the coordinates of my Humanoids's forward heal after he strides
> forward. During Cross-Over pose to come in 3 Key-Frames, he moves his
> back leg
> next to his forward leg. This move shifts his weight over forward leg, as
> his
> centroid translates to be directly vertically over his forward heal.
>
> Having said all that, throughout 3 key-Frames, I donot want his forward
> foot to
> move at all as the joint rotations of some other 10-11 joints might be
> changing, including the immediate parents of forward_heal_joint &
> forward_foot
> pair's parents, which are forward leg's knee, and hip joints.
I think this is where you need to be able to change which body part is
fixed. In my system I just specify the new fixed body part as being the
forward foot. From then on, any small spurious rotations affect the back
foot position rather than shifting the front foot.
> They get rotated
> to bring Humanoid_Root over the forward heal. They are slightly bent as
> they are
> weight bearing.
>
> I was hoping for some constraint specification that would keep an object
> location wise gripped.
Having a single fixed body part obviously contrains that body part, but when
you get into fixing multiple body parts (e.g. on stairs where both feet can
be in contact with the steps and both hands on the handrails for a number of
animation frames) then it starts to get an order of magnitude more complex.
This gets you into inverse kinematics which I never got round to
implementing properly in POV-Person, although I did have some thoughts on
how I'd go about doing it.
There's no single solution to inverse kinematics as the body parts could
more or less take any route from one fixed position to another. Also this
requires some way of dealing with situations where two body parts can no
longer be joined without violating other constraints. POV-Person was written
to allow for this to a limited extent in that it calculates a pose by
scanning the arrays back and forth rather than working from a single fixed
starting point. If two fixed positions are set and they can't be joined
it'll work out what it can and just leave a gap between body parts it can no
longer join. The next logical step would be to track back and 'straighten'
nearby joints to attempt to bridge the gap.
>
> Another application of constraint specification could be to ensure that
> the
> floor my exercise room is solid, so as I move my Humanoid around, I would
> not
> be able to dip any part of him/her below the surface.
>
This brings you into collision detection which is another somewhat complex
area of study. Once again, I wrote POV-Ray to be able to respond to such
detection by maintaining body part positions and rotations in arrays that
can be adjusted iteratively, but never implemented very much in that
respect. My forays into this area were largely contained in the crowd
positioning macros, which use the POV-Ray trace command to detect suitable
positions on a surface for a person to stand on. The macros actually shoot
out multiple traces for a given position to check the gradient of the
surface and the proximity of nearby vertical dropoffs and objects.
> Well, s/he has been entertaining me emerging through walls, stepping into
> floors. Now just by a new comers luck and mistakes, I had learned how I
> can
> dismantle him/her, segment by segment into a pile. No, no I am not into
> violence, but animating what happens to a being when hit by a car could be
> doable, not very realistically, for now. Given how little I had
> accomplished
> for having worked as long as I had on rudimentary animation, and how much
> I had
> discovered I don't know as of yet.
>
Well human animation is one of the most difficult things to get right. If it
were an easy subject then the professional animators who've been struggling
with it for years would have it nailed by now. There have been a couple of
recent films where, seen at a distance, you can't really tell which of the
figures are real or generated, but I've not yet seen anything that's very
convincing up close.
Our brains are extremely sensitive to even the very slightest inconsistency
in human appearance, posture and movement. Despite being completely
surrounded by images and representations of humans, it's very rare that we
would mistake any of that for a real person for very long. Even when someone
jumps and says "I though that was a real person" it's usually where they've
seen it out of the corner of their eye and the deception only lasts a tenth
of a second or so.
Regards,
Chris B.
Post a reply to this message
|
|