Digital Holographic Microscopy with Deep Machine Learning for Automated Determination of Dispersed Phase Parameters of Materials
N. A. Kuzmin, Yu. D. Arapov, A. E. Dormidonov, V. G. Kamenev, D. E. Ergashev, P. N. Yaroshchuk
Dukhov Automatics Research Institute, Moscow, Russia
Keywords: holography, machine learning, shock wave loading, dispersed phase
Abstract
A method for automating the determination of dispersed phase parameters in materials is proposed, based on digital holographic recording and deep machine learning. Numerical and field experiments were conducted on holographic recording of a cloud of particles with sizes on the order of 10 µm, and the resulting holograms were reconstructed layer-by-layer. It was demonstrated that a neural network is capable of recognizing particles in reconstructed holograms with accuracy metrics exceeding 75%, which is more than twice the accuracy of binarization and the Hough, Sobel, and Canny algorithms.
|
