Building Occupancy Counting from Video Preserving Occupants' Privacy

Abstract

Occupancy counting in buildings is important for a variety of reasons, including resource al-location in smart buildings and emergency response. Because most individuals spend more than 90% of their time inside a structure, creating a welcoming and relaxing atmosphere is critical. The traditional HVAC system regulates the rooms under the assumption of maximum occupancy in order to assure comfort. As a result, in the United States, the HVAC system accounts for more than half of all building energy use. In order to minimize energy waste, occupancy-controlled heating, ventilation, and air conditioning (HVAC) systems monitor how many people are in a given space and adjust the temperature and humidity accordingly. Occupancy models with real-time estimation and detection of building occupancy have been developed, and attention has been called to methods that leverage sensors that have already been installed in the buildings for reasons other than occupancy estimation. The high precision of camera-based occupancy estimation systems has made them popular. Privacy is the primary concern when employing a camera to count the number of occupants. In the prior methods, the privacy of the occupants was not addressed. The video frames have been blurred to protect the privacy of the occupants in this study and various occupancy counting methods were implemented, some of which do not necessitate deblurring while others do. We proposed a new approach combining statistical single-image blind deblurring and density estimation for calculating the number of occupants in a building based on the data collected from blurred surveillance video while yet maintaining some privacy for the building’s occupants. This approach has an error rate of 16.29%, which is lower than the other methods.