Estimation of occupancy level in indoor environment based on heterogeneous information fusion

Monitoring the number of occupants in each zone of a building is important for energy-efficient control of the HVAC system and the lighting system under normal conditions and for fast evacuation under emergency conditions. There usually exist multiple systems for localizing and monitoring the occupants in a building such as the active RFID system and the video cameras. The accuracy of each system is affected by different factors. Further hardware investment is usually required to improve the accuracy of each system. It is thus of practical interest to combine multiple systems to achieve higher counting accuracy without further hardware investment. In this paper, this problem is formulated as an information fusion problem under the criterion of minimum mean square error. However, it is usually difficult to solve the problem optimally due to the lack of data on the joint distribution of the observation noises of multiple systems. Two approximation methods are developed following the independence assumption and heuristics, respectively. Experimental results show that the two methods improve the accuracy of the active RFID system and the video cameras by around 43% and 73%, respectively.