In the sprawling, interconnected world of smart buildings, the ability to track and understand human movement indoors has long been a holy grail. Imagine a therapeutic space, say a large rehabilitation center, where every movement counts, and every step towards recovery needs to be monitored and analyzed. This is the realm where Hyeokhyen Kwon, from the Department of Biomedical Informatics at Emory University, has made significant strides. Kwon’s recent study, published in the IEEE Journal of Indoor and Seamless Positioning and Navigation, delves into the feasibility of indoor localization and multiperson tracking using a sparsely distributed camera network with edge computing.
The study, which spans a vast indoor area of 1,700 square meters, employs an end-to-end edge computing pipeline. This pipeline utilizes multiple cameras equipped with tensor processing units (TPUs) to achieve localization, body orientation estimation, and tracking of multiple individuals. The system’s privacy-preserving design is a standout feature, ensuring that while movements are tracked, individual identities remain anonymous. “Our approach focuses on extracting poses and bounding boxes, which are crucial for indoor localization and body orientation estimation,” Kwon explains. “This allows us to maintain a high level of privacy while still gathering valuable data.”
The implications for the energy sector are profound. Smart buildings are increasingly becoming the norm, and the ability to monitor and optimize human movement within these spaces can lead to significant energy savings. For instance, understanding how people move through a building can help in designing more efficient HVAC systems, lighting controls, and even space utilization. “By tracking movement patterns, we can identify areas that are frequently used and those that are not,” Kwon notes. “This information can be used to optimize energy usage, reducing costs and environmental impact.”
The study’s results are impressive: an average localization error of 1.41 meters, a multiple-object tracking accuracy score of 88.6%, and a mean absolute body orientation error of 29 degrees. These metrics indicate that the system is not only feasible but also highly accurate, even in a large indoor space with privacy constraints.
The commercial impact of this research could be transformative. Imagine a future where every large building, from hospitals to corporate offices, is equipped with such a system. Energy consumption could be drastically reduced, and operational efficiencies could be significantly improved. The potential for cost savings and environmental benefits is enormous.
As we look to the future, the integration of edge computing and computer vision technologies in indoor spaces is set to revolutionize how we interact with and manage our built environments. Kwon’s work, published in the IEEE Journal of Indoor and Seamless Positioning and Navigation, is a significant step forward in this direction. It paves the way for more intelligent, efficient, and privacy-conscious indoor spaces, shaping the future of smart buildings and the energy sector.
