+ 55 11 3091-5332

Take a look at our publications covering renewable energy, read what our authors and editors say on the topic, and browse selected research Publications

Automatic Selection Of Reference Lines For Spectrometer Calibration With Recurrent Neural Networks
Angel Zenteno, Adriana Orellana, Alex Villazón & Omar Ormachea

Instrument calibration is a critical but time-consuming process in many scientific fields. In this paper, we present an approach using recurrent neural networks (RNNs) for automatically detecting reference lines required to calibrate a spectrometer with well-known wavelengths of mercury and neon spectra. RNNs are a type of neural network that is best suited for processing sequential data. We collect a dataset of spectral images by taking images with cameras of different resolutions to train the neural network. Moreover, we prove that RNNs can learn to predict spectra lines in the calibration process with high precision. We match spectrometer measurements to their corresponding wavelengths by fitting a polynomial with these predicted reference lines. We validate our method using a 3D-printed spectrometer and compare the results with the NIST Atomic Spectra Database. The automatic selection of neon or mercury reference lines helps the calibration procedure to become faster, thus avoiding any manual selection. Our proposed technique is suitable for spectrometry applications where the speed is critical and the calibration process needs to be performed frequently.

Web Link
Read More
Published Date
Add by
Omar Ormachea