INERTIA

Occupancy Extraction, Modelling and Prediction in Tertiary buildings

The following video presents OCCUPANCY EXTRACTION using sensors (in this case Kinect cameras) from tertiary buildings, used for MODELLING and ultimately PREDICTION of building occupancy, which was developed in the scope of the INERTIA FP7 funded project. Based on real-time occupancy information per space/zone, as produced by the INERTIA Occupancy Extraction mechanism, as well as on historical data, the application can perform short-term (within a few minutes or hours from current time) and mid-term (intra-day or for the whole next day) prediction about the level of occupancy (empty, low-occupied, medium-occupied, nearly full, full) in each building space. The accuracy of the prediction result calculated by the Occupancy Prediction mechanism is compared against the ground truth data, as well as against the historical average values and the result that would be given by using the best matching relevant Open Reference Model. The results reveal that in most cases the prediction captures sufficiently ground truth and outperforms the historical average and the Open Reference Model. The scenario that is displayed is a real-case scenario referring to two different spaces of CERTH-ITI premises, used to extract real occupancy data, collected using Depth-Image cameras (Kinect). Some of the collected data were used for training the model, while the rest were left out as ground truth for testing purposes.

 

INERTIA System video demonstration