A.V. Bogoslovsky1, A.S. Afanasiev2, I.V. Zhigulina3, M.A. Pantyukhin4

1-4 MESC AF “N.E. Zhukovsky and Y.A. Gagarin Air Force Academy” (Voronezh, Russia)

1 p-digim@mail.ru; 2 A.S.Afanasiev1990@yandex.ru; 3 ira_zhigulina@mail.ru, 4 pull-request@mail.ru

One of the areas of video information preprocessing in machine vision systems is the selection of image areas whose shape and size correspond to the objects of interest. The use of two-dimensional discrete filtering of the Wiener-Hopf type for this purpose is associated with the impossibility of constructing an optimal filter and the problem of determining the filter aperture size, while an inaccurate selection of a prototype object can lead to "missed target" errors. These problems can be solved by introducing hierarchical preprocessing consisting of at least two stages. The use of two-dimensional discrete filtering at the first stage should ensure a sufficiently high probability of false alarm with a minimum probability of missing a target.

The purpose of the work is to study the capabilities of quasi-optimal two-dimensional discrete filtering of the Wiener-Hopf type for solving object search problems at the first stage of hierarchical image preprocessing.

The paper shows that to synthesize a discrete filter, it is necessary to use an image of a prototype object on a certain background. The concept of a "prototype object window" with an arbitrarily specified contrast is introduced. The paper considers the processing of test images by filters synthesized based on the window of a prototype object and having apertures of the smallest possible sizes, which reduces the influence of nearby objects on the filtering results. All possible combinations of contrast signs of the object of interest and the prototype object are considered, and their influence on the formation of artifacts in the processed image is analyzed. The areas of artifacts that are formed inside and around objects due to the influence of both object samples and background samples are determined. To take into account the filter response in these areas, it is proposed to use a decision window whose parameters are determined by the filter impulse response, the size and shape of the object of interest. It is found that threshold processing based on the maximum output signal level does not allow one to select similar objects that differ in size, so a new criterion, "maximum sum of video signal sample modules in the decision window," is proposed. The results of applying the criterion when processing with the smallest possible filter apertures are presented. The use of the maximum criterion of the sum of the modules of video signal samples in the decision window allows implementing the first stage of hierarchical image processing, where errors of missing targets are minimized and potential areas containing objects of interest are identified. In turn, this helps to minimize computational costs and the probability of false alarms at the second stage of hierarchical processing. The results obtained can be used to automate the processing of video information in machine vision systems.

Bogoslovsky A.V., Afanasiev A.S., Zhigulina I.V., Pantyukhin M.A. Decision window and criterion for its use in two-dimensional discrete filtering. Radiotekhnika. 2024. V. 88. № 11. P. 78−87. DOI: https://doi.org/10.18127/j00338486-202411-11 (In Russian)

1. Remington L. Klinicheskaja anatomija i fiziologija zritel'noj sistemy: Per. s angl. M.: ID Gorodec. 2020. 336 s. (in Russian).
2. H'jubel D. Glaz, mozg, zrenie: Per. s angl. M.: Mir. 1990. 239 s. (in Russian).
3. Changizi M. Revoljucija v zrenii: chto, kak i pochemu my vidim na samom dele. M: AST. 2015. 304 s. (in Russian).
4. Djevis R., Tverk M. Komp'ternoe zrenie. Sovremennye metody i perspektivy razvitija. M.: DMK Press. 2022. 690 s. (in Russian).
5. Marr B., Uord M. Iskusstvennyj intellekt na praktike. M.: Mann, Ivanov i Ferber. 2020. 320 s. (in Russian).
6. Egorova E.V., Aksyaitov M.H., Ryibakov A.N. Veroyatnostnaya protsedura raspoznavaniya i identifikatsiya prostranstvennyih obyektov. Uspehi sovremennoy radioelektroniki. 2023. № 3. S. 53–63. DOI: https://doi.org/10.18127/j20700784-202303-05 (in Russian).
7. Favorskaya M.N., Petuhov N.YU. Raspoznavanie prirodnyih obyektov na aerofotosnimkah s primeneniem neyronnyih setey. Avtometriya. 2011. № 3. S. 34–40. (in Russian).
8. Syrjamkin V.I. i dr. Adaptivnye sistemy tehnicheskogo zrenija. M.: RUSAJNS. 2024. 448 s. (in Russian).
9. Gonsales R., Vuds R. Cifrovaja obrabotka izobrazhenij. Izd. 3-e, ispr. i dop. M.: Tehnosfera, 2019. 1104 s. (in Russian).
10. Obrabotka mnogomernyh signalov. V 2-h knigah. Kn.1. Linejnaja mnogomernaja diskretnaja obrabotka signalov. Metody analiza i sinteza. Pod red. A.V. Bogoslovskogo. M.: Radiotehnika. 2013. 168 s. (in Russian).
11. Bogoslovskij A.V., Pantjuhin M.A., Zhigulina I.V. Algoritm vybora apertury fil'tra pri soglasovannoj so svojstvami vhodnogo izobrazhenija fil'tracii verhnih prostranstvennyh chastot. Radiotehnika. 2016. № 10. S. 128–132. (in Russian).
12. Pantjuhin M.A., Bogoslovskij A.V., Zhigulina I.V. Dvumernaja diskretnaja fil'tracija testovyh izobrazhenij. Radiotehnika. 2020. № 3(5). S. 64–72. DOI: 10.18127/j00338486-202003(05)-07 (in Russian).