Light has a significant impact on our behavior and the way we feel. Using
artificial light, we aim to substitute daylight in a time, when the amount of it is
insufficient, although not every type of lighting is suitable for our everyday needs,
living and working indoors.
With this masters thesis we aim to design a tool for controlling and measuring
the external (day)light rays, resulting in a copy of them in a controlled environment.
One of the ways to perceive colors and define their effect is called CCT – Correlated
Color Temperature. In order to do so we developed a system of sensors, which is able
to perform CCT value measurements and then send the information to the steering
system which in the end changes the CCT of the artificial lamps. Firstly, our system
transforms the raw data from the five RGB sensors into a XYZ color space, from which
we then get a CCT value reading. From this value reading we can then create artificial
light which successfully imitates its natural source. One of the most important parts of
our work when calculating the CCT is calibrating the sensors. Specifically, setting up
the transformation matrix from one color space to another. We did this using machine
learning.
The results of the evaluation of 1080 measurements (which were done using a
spectoradiometer) showed us, that using machine learning is a successful way to
calibrate the utility of low-cost RGB sensors.
|
