Occupied space in an elevator

Hi, I am thinking of the best way to measure the available space (in percentage) in an elevator with a camera mounted on the ceiling. There can be persons or objects (like trolleys, carts, etc.) inside.
A possible very simple idea (we don’t take image distortion due to top view into account):

  • detect the floor area when the elevator is empty (kind of initialization) and draw a rectangle on it
  • draw bounding boxes on foreground elements (people, objects, etc.)
  • do intersection of all the foreground bounding boxes with the floor area (to get the occupied space)

We don’t need classification, as what matters is the foreground objects (whatever they are).
Does OpenMV have any multiple bounding box model ?

We don’t have multiple bounding boxes yet. However, this is actually coming quite soon. In the mean-time frame differencing will work great for this.

How would I implement frame differencing? In a way, I should capture the baseline photo (no people inside) and then continuously compare with video stream ? What would be the proper functions to use?

There’s an example called frame differencing that show you how to do this.

1 Like

Hi
working on the frame differencing and looks promising. I am detecting blobs (when somebody enters the lift), draw rectangles around them and intersect these rectangles with the main floor of the elevator to calculate the percentage of fill area.
I was thinking of placing Apriltags at the 4 corners of the floor to initialize the floor area at camera side (and so draw that rectangle). Would it be a proper approach or something else instead?

Yes, that would work nicely.