Fast 3D particle reconstruction using a convolutional neural network: application to dusty plasmas. Issue 4 (2nd September 2021)

• Record Type: Journal Article
• Authors: Himpel, Michael; Melzer, André
• Is Part Of: Machine learning: science and technology. Volume 2:Issue 4(2021)
• Publication Date: 2021-09-02
• Subjects: 3D -- particle -- reconstruction -- neural -- networks -- vision -- dusty plasma; 006.31
• DOI: 10.1088/2632-2153/ac1fc8
• Languages: English
• ISSNs: 2632-2153

Abstract

Abstract: We present an algorithm to reconstruct the three-dimensional positions of particles in a dense cloud of particles in a dusty plasma using a convolutional neural network. The approach is found to be very fast and yields a relatively high accuracy. In this paper, we describe and examine the approach regarding the particle number and the reconstruction accuracy using synthetic data and experimental data. To show the applicability of the approach the 3D positions of particles in a dense dust cloud in a dusty plasma under weightlessness are reconstructed from stereoscopic camera images using the prescribed neural network.