Spectrometry for Light Bulb Classification

Abstract

Light bulbs energy consumption are paid by city halls to electrical companies in Brazil. These electrical companies do not have a guaranteed aspect of each light bulb's characteristics. Light bulbs with distinct power and steam type have different energy consumption. Most of the time, to guarantee the light bulbs characteristics a technician has to climb and verify each light pole. It is clear that this task is exhaustive and has to be repeated many times a year. In this work, a system is proposed capable of measuring the light bulb's characteristics easily. The system utilizes machine learning models to identify the required characteristics using as input the light spectrum. The results show that the trained models are able to identify correctly steam type, power, and brand.