Automatic Detection of Objects Blocking Elevator Doors using Computer Vision

In this paper we present a new approach applying computer vision methods to image data acquired with depth perception cameras to map the interior of the elevator, detect the position and the state of the door and to detect objects in the door area. The depth data is used to determine the elevator cabin as safety cube, i.e. the position of the door, layout of the elevator and so on, while color data further enhances the detection of new objects. The approach can detect the state of the elevator door as either opened or closed, while no object is blocking the view to the door, as well as successfully identify objects blocking an open door. This elevator monitoring proves to be relevant for determination of the elevator state, safety as well aspects of predictive maintenance.