Forum

Kinect - simple point cloud


#1

Hi people,

i am kind of “standing on a hose” - like the Germans say.

I wold like to have a simple point cloud and send the XYZ coordinates of each point somewhere else via udp.

I am using the VVVV-Kinect-1.0 plug in.

a push in the right direction would be great!

Tanks, Friedrich


#2

yea don’t do that ;]
320X240*3 values 30 fps via UDP you need 230400 opened ports and same amount of udp recivers to make it work hahahaha
just imagine patch with 230400 UDP recivers LOL


#3

You could get the XYZ Information using the Depth Texture node included with the openni-plugin and Pipet (EX9.Texture) . You´ll have to create the XY point-grid by yourself, just have a look at the pipet help-patch. The pipet will output a spread of colors, put these into HSL (Color Split) then feed the Lightness output into a Map (Value), correctly mapped this will give you the depth (Z). Put together XY and Z using Vector (3d Join) and put them into NetSend (Network Value). Voilà.

Depending on the desired resolution you may run into performance problems, pipet needs a lot of resources.

The 230400 ports stuff is nonesense.


#4

cool thx bjoren haven’t know that NetSend(Network Value) exists!


#5

Thanks for tips so far!

for context: http://www.grasshopper3d.com/video/kinect-grasshopper that is what i would like to do, but with mighty vvvv instead of processing. She is sending the data via udp too (a loooong coma separated string - i guess), but if there is a way to do it using less resources i would prefer it.


#6

hope this works, this needs to be done by a shader, i need to learn how to do it,

pointcloudbasec4.v4p (40.2 kB)


#7

guess only thing can help is cut-off unused points. think you will be ok on 1/3 of res


#8

still on my pointcloud with pipet the z values are not rigth, the formula its out there, but i need to use the expression node ? i dont get it


#9

@antokhio I could reduce the number of point in grasshopper.

All i need from vvvv is a bunch of points in space and a good way to send their coordinates to grasshopper. In grasshopper there is a tool which analyses the brightness of a still picture. Isn’t there something similar in vvvv: the brighter the pixel the higher the value? With that i could just put a matrix of points “over” the depth texture and evaluate the pixels under each point for its brightness. That value would be the Z-coordinate of that point. This would result in a very simple point cloud. <— strike that!

thats what Pipet (EX9.Texture) is for. thanks bjoern. i think i can take it from here. i will post my vvvv and grashopper solution when its done.


#10

@ffffriedrich did you check the file that i uploaded ? "pointcloudbasec4.v4p
"


#11

@andresc4, hi i’m very very amateur in the vvvv langage, and I download your “pointcloudbasec4.v4p” and when I open it in vvvv, I just have this ( look at the picture please), how I can see on the renderer the visions of my kinect v2?

My second question is how I can export the renderer in a sequence of .obj files to after import in a 3D software.

Thank for your answer and your future help.


#12

wow this is a old reply!
this is the way to go kinect2-nodes