Luminance Monitoring and Daylighting Control in Buildings Using Low-Cost Cameras

Michael Y Kim, Purdue University

Abstract

Daylighting has a significant impact on occupants, including not only visual comfort and visual task performance but also on workplace satisfaction and psychophysiological responses such as alertness, mood, and circadian rhythm. Intelligent, dynamic control of daylight via a building automation system is therefore crucial to maximizing the positive impacts of daylight provision in perimeter offices. However, existing daylight-linked controls (DLCs) lacks the fundamental ability to govern indoor luminous condition in a human-centric manner. It is mainly because the sensing technologies – such as ceiling-mounted photosensor - adopted in current DLCs are incapable of monitoring sufficient physical variables to suit such purpose. Hence, this Thesis aims to utilize a High Dynamic Range Imaging (HDRI) sensor in DLCs to unlock abilities to enhance human-centric features that have not been possible with conventional photosensors. The sensor, made of a low-cost programmable camera can capture a wide-area luminance distribution highly correlated with occupant visual perception, compared to conventional illuminance-based metrics. This Thesis begins with a development of a window-mounted HDRI sensor for real-time detection of potential glare sources including the sun. The sensor can capture the full luminance distribution of the exterior scene visible through the window and identify and locate potential sources of glare. To overcome the pixel-overflow by the extreme luminance of the sun and to estimate the accurate 3D position of the glare sources, the HDRI sensor was upgraded into a new fisheye-stereovision sensor made of dual cameras with different exposures. Experiments in fullscale offices showed that the calibrated window-mounted HDRI sensor can efficiently identify and locate potential glare sources in real time. The daylight control implementation included integration with shading controls to mitigate the risk of glare and comparison with conventional shading operation. Monitoring of indoor luminance distribution is equally important for human-centric DLCs. There are practical challenges in utilizing the HDRI sensor for monitoring luminance distribution perceived from the occupant perspective. Therefore, a new framework was developed for nonintrusive monitoring of luminance distribution perceived from occupant field-of-view (FOV), using a fisheye HDRI sensor installed at a non-intrusive position. The framework leverages the state-of-the-art photogrammetry (Structure-from-Motion – Multiview Stereo) pipeline to automatically reconstruct 3D surfaces of the room, which will be used for re-projection of luminance map captured by HDRI sensor into occupant FOV. To validate the performance of the framework, a systematic performance evaluation was conducted in a real-office experiment under variable lighting conditions to compare the reprojected luminance maps and the actual luminance measurement captured from occupant positions.

Degree

Ph.D.

Advisors

Konstantzos, Purdue University.

Subject Area

Design|Neurosciences

COinS