Strange phenomenon in depth. RRS feed

  • Question

  • I'm working on a MA thesis in which I use Kinect v2. And I have a strange phenomenon in depth.

    The presented images are recorded at a distance of 75 cm.

    Steel table and strange oblong black shapes in the image of depth, Direct Distance in black areas is
    0 mm, which means that the value of the depths went beyond the scope 500 <x <1200 mm.




    Wooden plank.


    The aluminum plate.


    The aluminum plate covered with foam rubber material


    Please  help me in finding the problem of the above-mentioned case.

    Friday, May 22, 2015 1:28 PM

All replies

  • Did you check if you have some interfering IR source?
    • Edited by jmmroldan Friday, May 22, 2015 2:06 PM
    Friday, May 22, 2015 2:03 PM
  • No I don't have any interfering IR source.

    I think the problem is with reflection.

    Friday, May 22, 2015 2:17 PM
  • It looks like the surfaces are all normal to the z axis, do these artifacts persist when you tilt the sensor relative to the surfaces? Perhaps you are getting too much specular reflectance back which is saturating parts of the IR sensor.
    Friday, May 22, 2015 2:45 PM

  • Normal view (the board is on the surface).

    Changed view (one side board raised).

    Friday, May 22, 2015 3:13 PM
  • In my experience, both reflection and absorption of the IR light has a primary affect on the Kinect v2 sensor.
    If the light is data.
    If the light is specularily tends to be oversaturated and the data for that pixel is invalid.

    The Kinect is a camera with a very bright IR light. Its similar to a bare lightbulb being held beside a digital camera and taking a picture. You will be hotspots where the digital camera is oversaturated due to the bare bulb, the angle of incident on the reflective material, and the material's reflective qualities.

    That is what I can generally seeing in your pictures. One was odd to me, the one labeled "The aluminum plate" and I would want to examine it in movement if I were exploring this issue.


    Monday, May 25, 2015 4:36 PM
  • It seems to me that what you wrote above, describes the problem well.On incorrect reading of data (black areas ), consists of 2things:
    1. The angle of 90 degrees to the observed surface.
    2. Material with a reflective surface.

    Because, when I change the angle of Kinect, field disappearor when I change the material on the rubber foam (light absorbing material).

    Thank you all for your help.

    Monday, May 25, 2015 5:54 PM