ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 4. Vol. 31. 2025

DOI: 10.17587/it.31.208-214

R. R. Diyazitdinov, Cand. Sc., Assistant Professor,
Samara State University of Railway Transport, Samara, 443083, Russian Federation

Binary Video Signal Multi-Parameter Classifier for Object Deformation Measurement

Received on 20.05.24
Accepted on 01.07.24

The article is devoted to binary video signal multi-parameter classifiers for object deformation measurement. The study examines an applied task. It is the rail deformation estimation of a railway track, which consists of several parts with different rail types. The problem is related to ensuring traffic safety on the railway. If the rail type is incorrectly recognized, the deformation parameters will be measured with a high error, which can lead to an incorrect assessment of the track condition and the omission of a potentially dangerous situation. The rail shape is measured using a machine vision system installed on a track measuring car. The signal of this system, representing a binary video signal(contour), is input to the classifier. Then the signal and the result of rail type recognition are used to estimate the deformation. The classifier meets the requirements to ensure real-time processing and high quality recognition. Therefore, the classifier takes into account the properties of the processed signals, on the one hand, simplifies processing, and on the other hand, ensures a low probability of classification error. One of the features of the classifier is its parametricity. The matching parameters were introduced into the classifier model: offset along the abscissa and ordinary axis, rotation angle. Parametricity allows to reduce the likelihood of recognition error. The classifier can be expanded to any number of types of rails, which makes it a universal solution for various roads, both Russian and European. The analysis of the measured deformation parameters showed that the proposed multi-parameter classifier provides a high accuracy (0.1 mm in the "wear side''parameter as the maximum absolute difference for the 0.95 level quantile). The proposed classifier can be used to estimate the condition of railway tracks.
Keywords: multi-parameter, classifier, recognition, deformation, binary video signal, track measuring car, rail type, machine vision

Acknowlegements: The research was done according to the state assignment of the Federal Agency for Railway Transport (Roszheldor) for researching, development and technological work for civil purposes Internet number / Registration number: 124040100033-0.

P. 208-214

Full text on eLIBRARY

References

1. Zheleznov D. V., Tarasov E. M., Isaycheva A. G., Mikheeva T. I. Development of the learning classifier of rail lines states with multivariate informative features, Trudi SPIIRAN, 2017, vol. 50, no. 1, pp. 32-54 (in Russian).
2. Tarasova A. E., Tarasov E. M. Automatic railway states classification, Vestnik transporta Povolg'ya, 2022, vol. 96, no. 6, pp. 79-83 (in Russian).
3. Klovskiy D. D. Discrete messages radio channels transmission, Moscow, Radio i svyaz', 1982, 304 p. (in Russian).
4. Proakis J. Digital communications, Moscow, Radio i svyaz', 2000, 800 p. (in Russian).
5. Instructcii, available at: https://tst.studyshark.ru/wp-content/uploads/2021/02/Èíñòðóêöèè - ïî - îöåíêå - ðê -¹436 ð -28.02.2020.pdf (date of access: 15.04.2024).
6. Metropoliteny, available at: https://www.metro.ru/library/metropoliteny/243/ (date of access: 15.04.2024).
7. DIN EN 13674-1-2011 Railway applications - Track -Rail - Part 1: Vignole railway rails 46 kg/m and above, available at: https://docs.cntd.ru/document/431971952 (date of access: 15.04.2024).
8. Orlov S. P., Efimushkin N. A., Efimushkina N. V. Deep neural network for diagnostics of railway track components, Vestnik Samarskogo gosudarstvennogo tehnicheskogo universiteta. Seriya: Tehnicheskie nauki, 2022, vol. 30, no. 1, pp. 63—74 (in Russian).
9. Faghih-Roohi S., Hajizadeh S., Nunez A., Babuska R., De Schutter B. Deep convolutional neural networks for detection of rail surface defects, Processing of the 2016 International Joint Conference on Neural Networks (IJCNN), 2016, pp. 2584—2589.
10. Pavlov V. A. Application of convolutional neural networks for detection and recognition of artificial space and ground objects, Radiotehnika, 2019, vol. 83, no. 12, pp. 58—67 (in Russian).
11. Diyazitdinov R. R. Iterative algorithm for accurate superposition of contours with non-uniform sampling step, Komp’yuternaya optika, 2023, vol. 47, no. 1, pp. 102—111 (in Russian).
12. EN 13848-1:2003 + A1:2008. Railway applications —Track — Track geometry quality — Part 1: Characterisation of track geometry, available at: https://standards.iteh.ai/catalog/standards/cen/3abf92b7-75e4-4b12-90de-af023dbf28c0/en-13848-1-2003a1-2008 (date of access: 15.04.2024).

To the contents