question about kinect_visionsensor filter dialog

[ljklonepiece]

Hi Marc,

I have been experimenting with the kinect sensor to obtain a depth map of the scene, but i am confused by the returned values in the "show filter dialog-->Extract coordinates from work image" whose return values are:

(1) point count along X
(2) point count along Y
(3) point1 x
(4) point1 y
(5) point1 z
(6) distance point 1
...

Originally I though x,y are coordinates in the x-y plane which are perpendicular to the direction of the infrared ray
and z should be the depth value. But after retrieving all the points of a scene, I found out x and z contains negative values where y is always positive, which means y is the depth value, is it?

I also want to know where is the origin of the plane, which is perpendicular to the infrared ray.
and what does distance point1 mean?

Thanks a lot in advance!

Yours sincerely,
Juekun

[coppelia]

Hello Juekun,

the coordinates that the Extract coordinates from work image filter returns are relative to the vision sensor's reference frame. If your vision sensor is looking at the scene like a camera would, with the head up, then:

• x would point to the left
• y would point up
• z would point to the front (i.e. in the viewing direction). Values should not be negative

Now if you need to further process those points, or if you want to display them, you will have to transform them via the vision sensor's transformation, in order to obtain coordinates in the absolute reference frame. In that case the z-coordinates could be negative, depending on the orientation of your vision sensor. But that transformation has to be done inside of a child script, by yourself.

The filter Extract coordinates from work image filter also returns with every coordinate a distance value, which is the distance between the vision sensor reference frame, and the detected point.

Have a look how the model Models/components/sensors/Hokuyo URG 04LX UG01_Fast.ttm works (check its child script, around line 38).

Cheers