Space utilization and activity recognition using 3D stereo vision camera inside an educational building

Abstract

Accurate occupancy information is imperative for the optimization of built-in environments to achieve energy savings and user comfort. Comprehending the occupancy information provides an opportunity to interpret movement patterns, circulation-flow, space usage patterns inside the building. In this paper, we designed a case study that includes experimental testbeds from the HL Linder Hall Cafeteria; a public shared space at the University of Cincinnati College of Business, United States. Based on the time-series data collected from 3D Stereo Vision Camera, an algorithm is proposed for the removal of redundant occupant IDs to overcome inconsistencies in the Field of View (FoV) of the camera and ensure accurate estimates and consistent data. Another algorithm for data annotation in activity recognition is proposed for the binary class classification of activity with sitting and moving labels. The data obtained can be used for inspecting various types of open and shared spaces available for work and quotidian interactions among occupants. The seats and space utilization patterns extracted from the camera within the monitored area are validated using a digitally advanced tool, known as ArcGIS Pro. For the experiment, prior permission was granted by the Building Management System (BMS) authorities, and occupants' confidentiality is preserved. The space usage patterns extracted can grant access to the new dimension of investigation associated with the space selection and occupant behavior inside the buildings. This paper also discusses the challenges faced during the design phase for the deployment, and it summarizes the potential improvements in the field of occupancy sensing for energy-efficient buildings.