Establishing an image-based ground truth for validation of sensor data-based room occupancy detection

Occupancy controlled heating, ventilation and air condition in offices are a mean to avoid energy waste due to over-conditioning of rooms. Occupancy detection methods based on sensor data that is already available in a building for other reasons than occupancy detection are therefore of great interest. However, such methods need to be validated according to a ground truth. This paper presents an image-based method for automatic establishment of a reliable and anonymous ground truth at entrance areas where the room height only allow the camera to be installed at 2.3-3.0 m above the floor. The method deploys various image processing techniques to infrared depthframe images recorded with a ceiling-mounted Kinect camera to anonymously detect and track persons entering or leaving a room and, consequently, counting the number of occupants in the room. The method demonstrated an accuracy of more than 98% in a stress test, 100% in a range of relevant function test and 99% in a three week long room occupancy test.