CLASSIFICATION OF IMAGES BASED ON AN ENSEMBLE OF STATISTICAL DISTRIBUTIONS BY CLASSES OF ETALONS FOR STRUCTURAL DESCRIPTION COMPONENTS
Keywords:Computer vision, methods of structural image recognition, set of key points, ORB descriptor, description components, statistical distribution, ensemble of solutions, classification efficiency criterion.
Context. Modern computer vision systems require effective classification solutions based on in-depth analysis of the nature of the data being processed. Statistical distributions are currently the primary means of analysis in image recognition systems. If the description of the recognized object is given by a large number of vectors, the statistical apparatus becomes a fundamental way to effectively decide on the class of the recognized object. This requires the use of a universal distribution apparatus in general for a system of multidimensional descriptions for established classes of data, defined by a given database of etalons. The classifier creates or organizes a new spatial structure of vectors from the elements of the analyzed object, which generally has some estimated similarity to the structure or composition of the etalon elements, and the classification is done by optimizing the degree of this similarity on the set of etalons. The probabilistic model of data generation is a key practical approach to formalizing the task of classifier training, the essence of which is to establish statistical distributions of objects or their components, followed by the procedure of aggregation of component solutions and further optimization in the environment of etalon classes. It is also valuable to study and apply criteria for evaluating the effectiveness in classification problem based on statistical principles.
Objective. Development of a method of effective classification of images by introduction of ensemble statistical decisions for structure of components of the description.
Method. A method for classifying images based on the construction of a generalized solution of an ensemble of components for which statistical distributions by data classes are preliminarily calculated is proposed.
Results. The synthesis of the classification method by applying the ensemble solution of the components of the description is carried out. The efficiency and effectiveness of the developed classifier are confirmed. On the examples of application of the method for synthesized data using traditional criteria, its effectiveness was experimentally evaluated.
Conclusions. The investigated methods of constructing an image classifier are based on an ensemble of partial solutions of statistical analysis data for the components of the structural description in the form of a set of key point descriptors. The statistical approach provides identification of the priority classification decision for components of the description on which set the resulting decision of ensemble is formed.
The scientific novelty of the study is the development of image classification method based on an ensemble of solutions of the component description, based on their statistical distributions by data classes.
The practical significance of the work lies in confirming the efficiency and effectiveness of the proposed methods on demonstration examples.
Duda R. O., Hart P. E., Stork D. G. Pattern classification, 2ed. Wiley, 2000, 738 p.
Gorokhovatskyi V. O., Gadetska S. V. Statistical processing and data mining in structural image classification methods (monohrafiia). Kharkiv, FOP Panov A. N., 2020, 128 p. DOI: 10.30837/978-617-7859-69-6.
Flakh P. Mashynnoe obuchenye. Nauka y yskusstvo postroenyia alhorytmov, kotorye yzvlekaiut znanyia yz dannykh. Moscow, DMK Press, 2015, 400 p.
Kim S., Kweon I.-S. Biologically motivated perceptual feature: Generalized robust invariant feature, Asian Conf. of Comp. Vision (ACCV-06), 2006, pp. 305–314.
Grosse K., Manoharan P., Papernot N. et al On the (statistical) detection of adversarial examples. arXiv preprint arXiv:1702.06280, 2017.
Kosmala A., Rigoll G. On-line handwritten formula recognition using statistical methods in Proceedings, Fourteenth International Conference on Pattern Recognition (Cat. No. 98EX170), Vol. 2, IEEE, 1998, pp. 1306–1308.
Filatov V., Kovalenko A. Fuzzy Systems in Data Mining Tasks, Studies in Computational Intelligence, 2019, pp. 243–274. DOI:10.1007/978-3-030-35480-0_6
Gorokhovatskyi V. O., Ponomarenko R. P. Klasyfikatsiia zobrazhen na pidstavi formuvannia nezalezhnoi systemy klasteriv u skladi strukturnykh opysiv bazy etaloniv, Suchasni informatsiini systemy, 2020, Vol. 4, No. 2, pp. 17– 23. DOI: 10.20998/2522-9052.2020.2.04.
Szeliski R. Computer Vision: Algorithms and Applications. London, Springer, 2010, 979 p.
Kohonen T. Self-Organizing Maps. Berlin, Heidelberg, Springer-Verlag, 2001. DOI:book/10.5555/558021
Paklyn N. B., Oreshkov V. Y. Byznes-analytyka: ot dannykh k znanyiam: ucheb. posob. SPb., Pyter, 2013, 704 p.
Gorokhovatskyi V. O., Pupchenko D. V., Solodchenko K. H. Analiz vlastyvostei, kharakterystyk ta rezultativ za-stosuvannia novitnikh detektoriv dlia vyznachennia osoblyvykh tochok zobrazhennia, Systemy upravlinnia, navihatsii ta zviazku, 2018, №1 (47), pp. 93–98.
Oliinyk A., Subbotin S., Lovkin V. et al. A The System of Criteria for Feature Informativeness Estimation in Pattern Recognition, Radio Electronics, Computer Science, Control, 2017, No. 4, pp. 85–96. DOI: 10.15588/1607-3274-2017-410
Gorokhovatskyi V. O., Gadetska S. V., Stiahlyk N. I. Vyvchennia statystychnykh vlastyvostei modeli blochnoho podannia dlia mnozhyny deskryptoriv kliuchovykh tochok zobrazhen, Radio Electronics, Computer Science, Control, 2019, No. 2, pp. 100–107. DOI: 10.15588/1607-3274-20192-11.
Nong Ye. Data Mining: Theories, Algorithms, and Examples (1st. ed.). CRC Press, Inc., USA, 2013.
Kacprzyk J., Pedrycz W. Springer Handbook of Computational Intelligence. Berlin Heidelberg, Springer-Verlag, 2015.
Muja M., Lowe D. G. Fast Matching of Binary Features, Conference on Computer and Robot Vision, 2012, pp. 404– 410. DOI: 10.1109/CRV.2012.60
Gayathiri P., Punithavalli M. Partial Fingerprint Recognition of Feature Extraction and Improving Accelerated KAZE Feature Matching Algorithm, International Journal of Innovative Technology and Exploring Engineering (IJITEE), 2019, Volume-8, Issue-10, pp. 3685–3690. DOI: 10.35940/ijitee.J9653.0881019
Haken H. Information and Self-Organization / H. Haken. – Berlin, Heidelberg: Springer, 2006.
Gadetska S. V., Gorokhovatskyi V. O., Stiahlyk N. I. Vyvchennia kryteriiv informatyvnosti danykh pry vprovadzhenni aparatu derev rishen u metodakh strukturnoi klasyfikatsii zobrazhen, Radio Electronics, Computer Science, Control, 2020, No. 3, pp. 78–87. DOI: 10.15588/1607-32742020-3-7.
Ethan Rublee, Rabaud Vincent, Konoligeet Kurt et al. ORB: an efficient alternative to SIFT or SURF, Computer Vision (ICCV), IEEE International Conference on IEEE, 2011, pp. 2564–2571.
How to Cite
Copyright (c) 2020 V. A. Gorokhovatsky, S. V. Gadetska, N. I. Stiahlyk, N. V. Vlasenko
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Creative Commons Licensing Notifications in the Copyright Notices
The journal allows the authors to hold the copyright without restrictions and to retain publishing rights without restrictions.
The journal allows readers to read, download, copy, distribute, print, search, or link to the full texts of its articles.
The journal allows to reuse and remixing of its content, in accordance with a Creative Commons license СС BY -SA.
Authors who publish with this journal agree to the following terms:
Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License CC BY-SA that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work.