Real-time particle concentration measurement from a hologram by deep learning

Abstract

Abstract Although the existing digital holographic technologies are effective in measuring particle concentration, the processes are cumbersome and time-consuming. The purpose of this study is to quickly and accurately measure the particle number from a single hologram through deep learning. The simulation and experimental results show that the prediction number are close to the ground truth without the complicated reconstructions and denoising processes, and the average relative error remains less than 10%. The prediction time of a hologram is at the millisecond level, which offers a new possibility for real-time processing.