View my account

Depth data at around 1m looks uneven

Comments

4 comments

  • MartyG

    Hi Nitin  It appears that the depth values in the lower half of the board are okay as they have colors that represent depth values above 0.  In the upper half of the board, the values are below zero.  As a correct depth value is 0 or higher, minus values likely represent an area of invalid depth values where the camera could not read depth detail from the board.

     

    Were you standing behind the board and holding the camera in your hand at a distance of 1 meter, please?  If you did then this might have cast a shadow from your body over the upper half of the board that could have made it more difficult for the camera to analyze for depth information.

    0
    Comment actions Permalink
  • Nitin

    Hi MartyG, 

    I doubt that shadow is the culprit here. The checkerboard is on a flat surface on a table and the camera is at one meter above it as shown in the picture below.

     

    0
    Comment actions Permalink
  • MartyG

    Checkerboards are typically used for camera calibration rather than measuring depth values from the board.  The depth sensing algorithm of RealSense cameras can become confused by repeating horizontal and vertical patterns such as tiled floors or ceilings, a phenomenon called Repetitive Pattern.  An Intel guide in PDF document format at the link below provides advice for negating the negative effects of such patterns.

    https://dev.intelrealsense.com/docs/mitigate-repetitive-pattern-effect-stereo-depth-cameras

    0
    Comment actions Permalink
  • Nitin

    Thanks MartyG. That's very helpful and good to know. 

    0
    Comment actions Permalink

Please sign in to leave a comment.