An Approach to Indoor Occupant Counting Method Based on Zenithal Video Analysis
Cuicui Wang, Ping Wang, Bo Song, Yue Wang, Zhenya Zhang · 2020
For many applications in buildings, such as heating, ventilation and air conditioning (HVAC), lighting control under normal conditions and fast evacuation under emergency conditions, the real-time and reliable knowledge of the number of occupants in each area plays an important role. This paper works on a naive and ideal occupant estimation model by detecting the enter and leave occupants of boundary detection area, and the zenithal video analysis is utilized to reduce the cumulative error of the naive estimation model. Specifically, the Kalman Filter is utilized to track the targets that through the border of each area, and two key tunable parameters that are the matching distance threshold and scaling factor are designed and tested elaborately to reduce the estimation error. Experimental results show that the proposed zenithal video analysis based indoor occupant counting approach with the careful tuning of matching distance threshold and the scaling factor,can effectively reduce the discriminant error of occupants entering and exiting, achieving the estimation of the indoor occupant number with above 92% accuracy, the occupied or unoccupied state estimation for an area with more than 99% accuracy, and the staying time of indoor occupant with more than 99% accuracy.