Sorting out "near" values


#1

yellow,

i’m posting again on the same topic since i’m pretty much stuck and i could really use some help.

The installation has a camera on top and I’m using tbeta to get x,y coordinates for every person walking the room.

the problem is that i can’t fine tune the camera settings to get one single x,y value for each person. instead i’m getting 2 or more blob positions for each person.(lets call this a point cloud). How can i get one single x,y value for every pointcloud based on a threshold of distance.?

i patched something with the intersect(Spreads sets)
but is working correctly for up to two point clouds.

Hope i’m understood. please help (here is the patch)

thanks you

intersectclose.v4p (12.2 kB)


#2

kind of a brute force method, draw circles at the positions and track them with a contour node:

intersectclose2.v4p (12.5 kB)


#3

ok that is nice and clever because it also gives the center of the points and not just one of them.(though it uses 12% of my cpu).

thanks a lot and respect to the community.:)


#4

quite a hard nut. here another patching approach:
it calculates the distances between each points, if distance smaller bla then merge and take the average.
problem is a line of points, where the first and the last don’t fit into the threshhold… you’ll see

intersectclose_try.v4p (29.6 kB)


#5

wow woei.it would take me a month to patch that.
indeed it acts a little weird on some occasions but i’m still to figure out exactly how it works. i’m starting to think that maybe a unify(2d) node (or something) would be useful in tracking situation.
for example a multitouch where you can have one type of interactivity for the fingers and one for the whole hand.

anyway thanks a lot you guys for your time and efforts.

regards


#6

hi levy,

woei admitting “this being a hard nut” really makes me curious.
IMNSHO woei is one of the users with deepest experience of vvvv.

this at a first glance.

edited:
@ 2nd glance: as expected: not a newbie patch provided by woei…


#7

this one really got me.
another approach.

Group (2d).zip (7.8 kB)


#8

nice one!.. no text …


#9

cool solution, kalle


#10

thx, i couldn’t resist to do a 3d version also…

Group (3d).zip (8.1 kB)


#11

well, this went much better than i expected. I’ll keep asking asking so u guys can make the vvvvorld a better place ;)

cheers


#12

vvvvelcome and of course keep asking.

doesn’t matter if your question are not that inspiring than this one was…

;)

i remember users who only wanted to be guide to a way and didn’t want readymade solutions. but i can’t help myself sometimes…