Statistical methods for hydrobiont images clustering

Yu.E. Shishkin, A.N. Grekov

Institute of Natural and Technical Systems, RF, Sevastopol, Lenin St., 28

 E-mail: iurii.e.shishkin@gmail.com

DOI: 10.33075/2220-5861-2020-1-153-159

UDC 681.3

Abstract:

     The paper studies the effectiveness of applying statistical approach to solving the problem of images clustering of aquatic organisms. A logistic regression model is used for a small number of classes. The process of constructing a statistical model is demonstrated using real images of plankton as an example. The transformation of individual organisms images into sets of factor space signs and the construction of separating hyperplanes in it are carried out. An estimate of occurrence probability of the first and second kind of errors in the implementation of binary image clustering using a separating hyperplane is obtained.

     The task of automatic clustering and identification of the video stream of hydrobiont images in real time is not exhaustively and fully resolved. The search for a solution to the problem is complicated due to the characteristics of the subject area: a wide variety of species and morphological features of plankton, large intraclass diversity and relative interclass similarity. In the case when recognition occurs manually, the influence of the human factor affects large volumes of monotonous work. The search for a suitable mathematical model of the classifier will greatly simplify the implementation of a numerical assessment of aquatic ecosystems productivity and the amount of incoming energy. The article deals with a special case of the image clustering problem with a small number of clusters and statistically distinguishable sets of hydrobionts features. This assumption is valid for ecosystems with limited species diversity, for example, the Black and Azov Seas. The proposed model is the basis of intellectualization when deciding on the appropriateness of using statistical clustering methods for the specific problem under consideration.

Keywords: statistical clustering, EM algorithm, data mining, machine learning, hydrobionts, anomaly detection, logistic regression.

To quote, follow the DOI link and use the Actions-Cite option or copy:

[IEEE] Y. E. Shishkin and A. N. Grekov, “Statistical methods for hydrobiont images clustering,” Monitoring systems of environment, no. 1, pp. 153–159, Mar. 2020.

Full text in PDF(RUS)

LIST OF REFERENCES

  1. Stepanovskikh A.S. General Ecology, Moscow: Unity Dana, 2000. 510 p.
  2. Turner J.T. The importance of small planktonic copepods and their roles in pelagic marine food webs // Zoological Studies, vol. 43, № 2, 2004. P. 255–266.
  3. Thompson P.A. Plankton: a guide to their ecology and monitoring for water quality // Commonwealth Scientific and Industrial Research Organization, 2009. P. 7–8. DOI: 10.1071/9780643097131
  4. Shishkin I.E., Skatkov A.V. Information technology for the detection of anomalies in monitoring observations: a monograph. Simferopol: IT “ARIAL”, 2019. 368 p.
  5. Shishkin I.E., Grekov A.N. Image clustering methods for an automated video recorder and plankton analyzer // Integrated research of the oceans: materials of IV Russian. scientific conf. young scientists. 2019. P. 380–381.
  6. Inzartsev A.V., Pavin A.M., Lebedko O.A. Recognition and inspection of small-sized underwater objects using autonomous uninhabited underwater vehicles // Underwater research and robotics. 2016. №. 2 (22). P. 36–43.
  7. Kharinov M.V. A generalization of three approaches to optimal digital image segmentation // Transactions of SPIIRAS. 2013. №. 2 (25). P. 294–316.
  8. Belim S.V., Kutlunin P.E. Highlighting contours on images using the clustering algorithm // Computer Optics. 2015.Vol. 39. №. 1. P. 119–124.
  9. Matveev Iu.N. Fundamentals of systems theory and systems analysis, Tver: TSTU, 2007. 100 p.
  10. Shishkin I.E., Grekov A.N. Analysis of image clusterization methods for oceanographical equipment // 2018 International Russian Automation Conference (RusAutoCon), At Sochi, Russia, September, 2018. DOI: 10.1109/RUSAUTOCON.2018.8501756.
  11. Faillettaz R., Picheral M., Luo J.Y. Imperfect automatic image classification successfully describes plankton distribution patterns // Methods in Oceanography Vol. 15, 2016. P. 60–77.
  12. Yakovleva T.V. Applicability conditions for the statistical rice model and calculation of the parameters of the rice signal using the maximum likelihood method // Computer Research and Modeling. 2014.Vol. 6. №. 1. P. 13–25.
  13. Sirota A.A., Solomatin A.I., Voronova E.V. A two-stage algorithm for detecting and estimating the boundary of objects in images under conditions of additive noise and deforming distortions // Computer Optics. 2010.V. 34. №. 1. P. 109–117.
  14. Gdansky N.I., Krasheninnikov A.M. Separation of objects in multidimensional feature spaces using normal classifiers // Social Policy and Sociology. 2012. №. 3 (81). P. 202–211.
  15. Tuganbaev A.A. Higher mathematics. Functions of many variables, double and triple integrals M.: Flint, 2019. 228 р.

If you have found a spelling error, please, notify us by selecting that text and pressing Ctrl+Enter.

 78 total views,  2 views today

Translate »

Spelling error report

The following text will be sent to our editors: