Performance evaluation of non-intrusive luminance mapping towards human-centered daylighting control

Abstract

Monitoring luminance distributions, as perceived by the occupants, offers enormous possibilities for continuous and reliable sensing and human-centered daylighting control. Novel luminance monitoring, that utilizes a camera sensor at a non-intrusive position (distant from the occupants), can be achieved using HRDI sensing coupled with occupant-perceived luminance map re-projections. To overcome computational and practical challenges, this paper implements such a luminance monitoring framework in a real daylit office space and evaluates its performance using measured data. The processed re-projected luminance maps were compared against HDRI-measured luminance maps from two side-lit occupant positions. The results, presented in terms of structural similarity measures and visual comfort prediction and diagnostic performance, show that the developed method can predict and monitor luminance distributions in the occupant field of view with sufficient accuracy. A regional super-pixel-scale analysis was performed to further analyze the source of errors, by applying masks to four different regions of interest. The major source of error is the bright exterior scene visible through the window, as well as bright reflections inside the space, which should be generally avoided for glare protection. Overall, this paper presents an important step towards non-intrusive, human-centered daylighting control based on luminance mapping.