Computer vision to advance the sensing and control of built environment towards occupant-centric sustainable development: A critical review

Abstract

As urban development progresses, the built environment control has faced more critical challenges in improving energy efficiency, air quality, and environmental comfort. Occupant information (e.g., occupant status and behavior) sensing is a key but challenging aspect of built environmental control. Computer vision (CV) technology provides a new way for multi-dimensional information acquisition. However, a critical review is lacking in the cross-research area of CV and built environment control, particularly considering the technological advancements following the COVID-19 pandemic. This article reviews the latest advancements in the built environment from international sources, with a focus on the research frontier in four branches: ventilation and indoor air quality control, COVID-19 control, thermal environment control, and lighting control. Through critical comparisons and analyses, it demonstrates that CV technology can effectively sense highly dynamic built environments, which greatly enhances the data dimension, resolution and accuracy compared to existing sensing technologies. Reported data shows that CV technology achieved an average detection accuracy of about 95% for occupant-related information and 86% for comfort-related information. Effective methods to improve the accuracy include incorporating data fusion by using other sensors, upgrading algorithms, and improving the model training. Particularly, the COVID-19 pandemic has driven the development of mask detection and social distancing detection using CV. The challenges, future trends and potential applications are discussed. This study emphasizes the need for cross-field integration of CV and built environment to facilitate the sharing of cutting-edge techniques and knowledge, which will stimulate more innovations in the future.