Photometers use correction filters to adjust spectral responsivity of sensors so that the combined spectral responsivity approximates the responsivity of the human eye V(λ). However, the combination of these components is hardware based, and the quality of the photometer depends on this combination. We propose a meter that uses an RGB sensor, a LED and an artificial neural network that transforms the output of the sensor into luminous transmittance, without the need of a filter. The ANN was trained and validated with two different spectra datasets and generated results with error values below 3%. The methodology presents an option for a meter with calibration that depends only on a software. This allows the development of a low cost and compact photometer.

Date Published: 27 February 2019
PDF: 15 pages
Proc. SPIE 10913, Physics, Simulation, and Photonic Engineering of Photovoltaic Devices VIII, 109131I (27 February 2019); doi: 10.1117/12.2510397

Published in SPIE Proceedings Vol. 10913:
Physics, Simulation, and Photonic Engineering of Photovoltaic Devices VIII
Alexandre Freundlich; Laurent Lombez; Masakazu Sugiyama, Editor(s)