Kinect measure object (get dimensions)

Hi
im trying to develop an app to measure an object placed on the floor by placing my kinect on the ceiling looking down (see image 1).

i want to get the “height width and depth” of any object in front of the kinect, just like a measure tape but faster

i tried to develop it using the kinect2+ MS sdk + opencv(emgu) but i think my method was wrong and someone suggested me that i should use a pcl library but i couldnt find any for c# so maybe openframeworks could be the way to go
hope you guys can guide me,
moving from c# to openframeworks is not a problem.

this is what my c# app does (with errors) :

  • filter depth image by some constant ( filter out floor )
  • detect objects using opencv
  • get the corner of my detected object and convert to real world coordinate
  • get distance between world coordinate corners to get “object width and depth”
  • get distance between highgest corner position and floor to get “object heigth”

in other words i try to detect objects on 2d (flatten depth image) then convert to 3d and then measure.

“The problem” im having is when i place an object away from the center of my image i get huge error like 15cm on the "width or depth " because opencv encloses sides of my object as if it were part of the top surface (see image 2 ) so instead of getting the bounding box of the surface of the object im getting the top+visible side of the object.


any ideas ?

Hello, do you have source code available? Working on a similar project and have made some headway. Might be able to offer some suggestions…

-K

Hello, did you figure out the solution. I am working on similar problem using C# & OpenCV. Can you please share the source code.