Prof. Yasuhiro Takaya, Osaka University, Japan

Title 

Ghost Imaging Microscopy enhanced by deep learning

Abstract

We propose a novel sensitive imaging method based on the Ghost imaging (GI) of which imaging time is drastically improved by using deep learning. The GI is one of a single pixel imaging by correlation between illumination patterns and detected light intensities. Random patterned illuminations are irradiated onto a sample. Then intensities of interactive light between the light and the sample, such as scattering, transmittance, fluorescence and so on, is detected by a bucket detector. So, the GI has an advantage for faint light intensity. However, a quality of an image depends on the measurement times. For obtaining a clear image of the GI, it is necessary to measure too many times. Therefore it is difficult to catch up with a fast phenomenon because of many measurements with different illumination for correlation analysis. To overcome this problem, we have applied a deep learning technique for reducing numbers of measurement. In the matter of a deep learning, the proposed method deals with a convolutional neural network (CNN). In this work, we derived a possibility of light in each pixel. A block diagram of the CNN includes some layers about a convolution and activation function. Furthermore, we repeated the set at three times. Finally, the probability map has been derived from the CNN. The validity of the proposed method was verified by fundamental experiments as well as numerical simulations. As a result, we have developed 60 times faster than the conventional GI. Additionally, we have observed a moving microparticle in 0.08 sec.

Biography

Yasuhiro Takaya received PhD. degree in precision engineering from Hokkaido University in 1992. He is currently a Professor of Department of Mechanical Engineering, Graduate School of Engineering at Osaka University. 

His research ambition is to establish novel principles of fundamental measurement and machining in nano/micro manufacturing engineering and science based on three “N”, namely, Nanophotonics, Nonequilibrium mechanics and Nanomaterials. Research scope covers mainly three technological categories as follows; 

Nano-Metrology

Micro-probe for the nano-CMM based on the laser trapping technique

Radiation pressure controlled micro-displacement sensor

Measurement of micro-cracks distribution in grinding glass

On-machine tool measurement for cutting-edge profile of micro endmill

Nano-Machining

Cu-CMP using water-soluble Fullerenol slurry

Surface analysis of chemical polishing process based on Raman spectroscopy under surface plasmon excitation

Laser nano-machining technique using femto-pulse photonic nanojets

Nano-Assembly

In-situ measurement system for the particle growth in self-assembly process based on fluorescence polarization

Development of new functional device based on novel nano physics

Novel-Metrology 

Photon-Metrology based on quantum effect

AI-Metrology based on machine learning, deep learning methods etc. 

Polarization measurement by quantum optics

He is a member of JSPE, JSME, JSAT, ASPE and CIRP Fellow member. He is a winner of a lot of awards such as JSPE Award in 1997, 1999, 2004, 2008, György Striker Award at IMEKO World Congress XIV in 1997, Outstanding Paper Award given by Fanuc FA Foundation in 2006, 2009, IJAT Best Review Award in 2016, IJAT Best Paper Award in 2017 and more.