The Andrew and Erna Viterbi Faculty of Electrical & Computer Engineering
High density volumetric flow profiling by point spread function engineering via deep learning
Determination of fluid flow-field on the microscopic scale is an area of research of high interest for micro-reactors, mixers and various lab-on-a chip applications. Popular methods for measuring the velocity field of a flowing fluid are Particle Image Velocimetry (PIV), and the closely related Particle Tracking Velocimetry (PTV). These methods are based on inserting small visible particles (e.g. fluorescent beads) that serve as tracer particles to report on the flow of the fluid.
In many applications 3D flow determination is of interest, however imaging of flowing particles with a microscope suffers from the same standard depth-of-field limitations as other microscopy applications, due to particles going out of focus. Consequently, several techniques have been developed that adapt the method to characterization of 3D flow.
In this work, a new approach is proposed to overcome this limitation by combining PSF engineering with Convolutional Neural Networks (CNNs); namely, training a CNN to estimate the flow velocity in a captured PSF engineered series of images (2 or more). The training of a CNN aims to tackle the problem in an end-to-end fashion, going around intermediate steps that are usually used in classical approaches such as localization and tracking.
* M.Sc. Under the supervision of Prof. Anat Levin and Prof. Yoav Shechtman.
Zoom meeting: https://technion.zoom.us/j/98390344065
Sun 20 Jun 2021
Start Time: 14:00
End Time: 15:00
Zoom meeting | The Andrew And Erna Viterbi Faculty Of Electrical & Computer Engineering