none
About these two variables in SkeletonFrame: SkeletonFrame.FloorClipPlane and .NormalToGravity RRS feed

Answers

  • Are you sure that the Y value of FloorClipPlane is positive? I'd expect it to be negative, pointing downwards. The X,Y,Z values should be the unit vector pointing in the direction of the floor from the kinect camera (at position {0,0,0}), and the W value should be the distance, in meters, from the Kinect camera to the floor, along the unit vector specified as {X,Y,Z}.

    So, X,Y,Z value should already be normalized to unit length. Consider W value separate from normalized floor direction vector.

    Eddy


    I'm here to help
    Monday, October 10, 2011 7:38 PM

All replies

  • NormalToGravity is not actually supported in the Beta 1 Refresh release. FloorClipPlane should be initialized, however, as long as Kinect device can see enough of the floor to give a value for this with high confidence. Can you send a sample image of what your depth frame looks like so we can get a better idea of your setup?

    Thanks!
    Eddy


    I'm here to help
    Monday, October 3, 2011 11:24 PM
  • Hi Eddy! Thank you for your answer~ I'm waiting for you

    Yes, I got the value of floorClipPlane with images have enough floor pixels.

    Another question is.. why the left most 4 columns of the depth image is 0?


    Thanks,

     

    Chenyang


    Chenyang
    Monday, October 3, 2011 11:32 PM
  • Hmmm, I would definitely expect FloorClipPlane to be initialized in your application, with W value giving you distance to floor (in meters), whenever one or more SkeletonData object has its TrackingState field initialized to SkeletonTrackingState.Tracked. Can you share some of your code where you try to use SkeletonFrame.FloorClipPlane?

    About left side of depth image being all zeros, it is a known issue that you should work around if it's causing you problems.

    Eddy


    I'm here to help
    Tuesday, October 4, 2011 12:52 AM
  • Hi Eddy,

    I'm currently not using the values in FloorClipPlane yet, would you please explain the meanning of these values, plz?

    For example, what I get for the frame above is : W X Y Z  =  1.743904 0.006349928 0.9304801 -0.3662871;

     

    Thank you!

     

    Chenyang


    Chenyang
    Tuesday, October 4, 2011 5:03 PM
  • Hi Eddy

     

    can I use this vector[w,x,y,z] to calculate the distance from a given position(like head) to the floorClipPlane? Why the result turns out to be weired? Is the position vector [1,x,y,z] need normalization?

     

    Thank you


    Chenyang
    Wednesday, October 5, 2011 6:08 PM
  • Are you sure that the Y value of FloorClipPlane is positive? I'd expect it to be negative, pointing downwards. The X,Y,Z values should be the unit vector pointing in the direction of the floor from the kinect camera (at position {0,0,0}), and the W value should be the distance, in meters, from the Kinect camera to the floor, along the unit vector specified as {X,Y,Z}.

    So, X,Y,Z value should already be normalized to unit length. Consider W value separate from normalized floor direction vector.

    Eddy


    I'm here to help
    Monday, October 10, 2011 7:38 PM
  • Thank you Eddy. How about the accuracy? It the accuracy decrease with the depth goes deeper? If I measure, say, height of a walking person?

     

    Thank you


    Chenyang
    Monday, October 10, 2011 7:42 PM
  • Yes, accuracy decreases a bit as you get farther from camera. For objects 1 meter away, accuracy should be within 1-2 centimeters, but objects 3-4 meters away could have accuracy within about 5cm.

    Eddy


    I'm here to help
    Monday, October 10, 2011 7:57 PM