Rapid Determination of Temperature Distribution in Diamond Anvil Cells from Multispectral Camera Images by Fully Connected Neural Network

Remote non–contact temperature measurement is an important task that is being solved both in space research and in the study of matter in extreme states in diamond anvil cells. The common approach is the least squares method for the deviation of the measured spectrum and the calculated curve of the spectral density of energy brightness. To determine the temperature and emissivity with minimal uncertainties, a hybrid method was proposed that allows minimizing only one parameter – temperature. In this case, it is possible to determine not only the emissivity, but also the coefficients of the polynomial decomposition of the emissivity. The solving performance problems when processing a large number of image points using neural networks is discussed in this paper.