Deep learning for nano-sized particle detection: initial feasibility check
other
One of the main challenges in detecting a nano-sized particle on a ‘rough’ surface is to differentiate the particle from speckle that is caused by the substrate. The ‘traditional’ approach in particle detection is to record multiple images under varying illuminations, and sum the recorded irradiance distributions. Since the speckles will be uncorrelated the corresponding irradiance modulations will average out. The particle will be (more or less) equally present in all images. Adding the various images thus improves the visibility of the particle. The ratio between the peak irradiance and the standard deviation of the summed image (peak-to-sigma, or P2σ) is a measure for the likelihood that a particle is present. Setting the P2σ threshold is a tradeoff between being able to detect particles of a certain size, whilst maintaining an acceptable false-positive rate.
TNO Identifier
867653
Publisher
TNO
Source title
SID Semicon Innovation Day, Science Centre Delft, 21 May 2019
Collation
1 p.
Place of publication
Delft
Files
To receive the publication files, please send an e-mail request to TNO Repository.