Human-centric light sensing and estimation from RGBD images: The invisible light switch

Lighting design in indoor environments is of primary im-portance for at least two reasons: 1) people should perceivean adequate light; 2) an effective lighting design meansconsistent energy saving. We present theInvisible LightSwitch(ILS) to address both aspects. ILS dynamically ad-justs the room illumination level to save energy while main-taining constant the light level perception of the users. Sothe energy saving is invisible to them. Our proposed ILSleverages a radiosity model to estimate the light level whichis perceived by a person within an indoor environment, tak-ing into account the person position and her/his viewingfrustum (head pose). ILS may therefore dim those lumi-naires, which are not seen by the user, resulting in an effec-tive energy saving, especially in large open offices (wherelight may otherwise be ON everywhere for a single per-son). To quantify the system performance, we have col-lected a new dataset where people wear luxmeter deviceswhile working in office rooms. The luxmeters measure theamount of light (in Lux) reaching the people gaze, whichwe consider a proxy to their illumination level perception.Our initial results are promising: in a room with 8 LED lu-minaires, the energy consumption in a day may be reducedfrom 18585 to 6206 watts with ILS (currently needing 1560watts for operations). While doing so, the drop in perceivedlighting decreases by just 200 lux, a value considered neg-ligible when the original illumination level is above 1200lux, as is normally the case in offices.