G.S. Baydin – Lecturer, Faculty of Computer Science and Management Systems, 

Bauman Moscow State Technical University (National Research University)

E-mail: baydin1015@gmail.com

A.S. Titov – Student, Faculty «Informatics and Management Systems»,  Bauman Moscow State Technical University (National Research University)

E-mail: toliakpurple@gmail.com

Nanoscale world exploration is increasing by researches. Nano measurement toolbox is continuously developing. Particle counting on a nanoscale image has wide range of application. With the increasing complexity of experiments with nanoscale images, the amount of data increases, therefore, there is a need for automation. The purpose is to develop a technique of automation particles counting on nanoscale image captured by electron microscope.

Defined, that preparatory image processing required to increase accuracy of recognition. Application of algorithm, called median nonlinear FIR filter is optimal.

Nanoscale images with particles, captured by electron microscope are the input data. Application of median nonlinear FIR filter is performed on the first stage of technique.

Segmentation of particles on nanoscale images is performed on the second stage of technique. Thresholding, Watershed, convolutional neural network U-Net and Spectral Clustering is considered for image processing to count the particles.

Thresholding algorithm description is performed. Threshold value has been used by the algorithm to complete segmentation. ISODATA, Mean, Triangle and Otsu threshold value determination methods description was provided. Advantages and disadvantages of each method was provided.

The following is a description of the Watershed algorithm. As a result of segmentation, the Watershed algorithm selects segments, each of which corresponds to a particle.

Then, a description of the U-Net algorithm is performed. The description of the architecture of the neural network U-Net. To train the neural network, images are used along with the resulting segmentation.

Finally, the Spectral Clustering clustering algorithm is described. As preparation image processing, Gaussian smoothing is used, which is necessary to obtain a uniform gradient. The algorithm consists of several steps: image smoothing; obtaining a weighted graph of the gradient connection; finding eigenvectors; and application of the k-means method.

For each algorithm, a description of the advantages and disadvantages is given.

A comparative analysis of the algorithms was performed. For comparative analysis, a test sample with many nanoscale images formed by an electron microscope was used. For a qualitative comparison of the proposed algorithms, the averaged and median data of the absolute and relative error were used.

When comparing the results, quantitative metrics were used. The following metrics were used: the range of variation of the relative error; and the variance of the relative error.

Based on the estimated results obtained, a conclusion is made on the qualitative and quantitative characteristics.

The output of the method for determining the number of particles in a nanoscale image of an electron microscope is: an image of particles with selected segmented particles; an image containing only a segmentation channel; and the number of particles in the image.

In conclusion, a comparison of the proposed tehnique with existing ones is given.

The given technique suggests one of the options for solving the problem of automated counting of the number of particles in a nanoscale image. As a result, the optimal sequence of various algorithms application is obtained.

1. Kobayasi N. Vvedenie v nanotekhnologiyu. M.: Binom. Laboratoriya znanij. 2008. 134 s. (In Russian).
2. Foster L. Nanotekhnologii. Nauka, innovacii i vozmozhnosti. M.: Tekhnosfera. 2008. 352 s. (In Russian).
3. Sojfer V.A., Kupriyanov A.V. Analiz i raspoznavanie nanomasshtabnyh izobrazhenij: tradicionnye podhody i novye postanovki zadach. Komp'yuternaya optika. 2011. T. 35. № 2. S. 136–144. (In Russian).
4. Podschet chastic v zhidkostyah. [Elektronnyj resurs] Rezhim dostupa: http://www.tehnopribor.ru/downloads/science-articles/schetchikchastits.php-clear_cache=Y (Data obrashcheniya: 20.12.2019) (In Russian).
5. American Meteorological Society Glossary. [Elektronnyj resurs] Particle. Rezhim dostupa: http://glossary.ametsoc.org/wiki/Particle (Data obrashcheniya: 15.12.2019)
6. Nacional'naya biblioteka MGTU im. N.E. Baumana [Elektronnyj resurs]. Mediannaya fil'traciya. Rezhim dostupa: https://ru.bmstu.wiki/Mediannaya_fil'traciya (Data obrashcheniya: 15.12.2019) (In Russian).
7. Ridler T.W., Calvard S. Picture thresholding using an iterative selection method. IEEE transactions on Systems, Man and Cybernetics. 1978. V. 8(8). P. 630 – 632.
8. Zack G.W., Rogers W.E., Latt S.A. Automatic Measurement of Sister Chromatid Exchange Frequency. Journal of Histochemistry and Cytochemistry. 1977. P. 25 (7). P. 741–753: DOI: 10.1177/25.7.70454.
9. Otsu N. A threshold selection method from gray-level histograms. IEEE transactions on Systems, Man and Cybernetics. 1979. V. 9. № 1. P. 62–66.
10. Bajdin G.S., Bukreev D.D. Metodika raspoznavaniya zlokachestvennyh novoobrazovanij v zheludke cheloveka na snimkah komp'yuternoj tomografii s ispol'zovaniem algoritmov obrabotki izobrazhenij. Biomedicinskaya radioelektronika. 2019. T. 22. № 5. S. 5–14. DOI: 10.18127/j15604136-201905-03 (In Russian).
11. Ronneberger O., Fischer P., Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. 2015. ArXiv, abs/1505.04597.
12. Haddad R.A., Akansu A.N. A Class of Fast Gaussian Binomial Filters for Speech and Image Processing. IEEE Transactions on Acoustics, Speech, and Signal Processing. 1991. V. 3. P. 723–727.
13. Zhang C., Zhu G., Chen M., Chen H., Wu C. Image Segmentation Based on Multiscale Fast. 2019. ArXiv, abs/1812.04816.
14. Yuan C., Yang H. Research on K-Value Selection Method of K-Means Clustering Algorithm. 2019. J. 2. 226–235. 10.3390/j2020016. 
15. Manifold System documentation. [Elektronnyj resurs] Images and channels. Rezhim dostupa: http://www.georeference.org/doc/manifold.htm #images_and_channels.htm (Data obrashcheniya 12.04.2020).
16. Kaspers A. Blob Detection. Biomedical Image Sciences. Image Sciences Institute. UMC Utrecht. 2011.
17. Ganyavin V.A., Bitus E.I., Saraev V.V., Manzevich A. Yu., Ganyavina N.A. Komp'yuternyj analiz mikrofotoizobrazhenij poristoj struktury netkanyh tekstil'nyh materialov. M.: MGUDT im. K.G. Razumovskogo. 2015. 158 s. (In Russian).
18. Astahov M.M., Kondar' V.I., Storozhuk O.M. Sposoby izucheniya mikrostruktur s pomoshch'yu opticheskogo cifrovogo mikroskopa. M.: MIFI. 2018. 124 s.(In Russian). 
19. Nikitin A.E. Postroenie analizatora izobrazhenij granulometricheskogo tipa. Nauchno-tekhnicheskij vestnik informacionnyh tekhnologij, mekhaniki i optiki. 2008. №46. Rezhim dostupa: https://cyberleninka.ru/article/n/postroenie-analizatora-izobrazheniy-granulometricheskogo-tipa (Data obrashcheniya: 10.02.2020) (In Russian).