Application of convolutional neural networks in the intelligence security system subsystem
The article proposes the structure of an integrated security system for areal facilities. The main subsystems included in its composition are considered. The tasks of the intelligence subsystem for detecting and recognizing ground-based objects of observation in a complex phono-target environment ar...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Belarusian National Technical University
2020-08-01
|
| Series: | Sistemnyj Analiz i Prikladnaâ Informatika |
| Subjects: | |
| Online Access: | https://sapi.bntu.by/jour/article/view/471 |
| _version_ | 1797873713229594624 |
|---|---|
| author | V. S. Demeshko A. I. Фёдоров |
| author_facet | V. S. Demeshko A. I. Фёдоров |
| author_sort | V. S. Demeshko |
| collection | DOAJ |
| description | The article proposes the structure of an integrated security system for areal facilities. The main subsystems included in its composition are considered. The tasks of the intelligence subsystem for detecting and recognizing ground-based objects of observation in a complex phono-target environment are defined.The task of detecting an object of observation was solved on the basis of a previously proposed algorithm. The disadvantage of this algorithm was the presence of false positives from a flickering complex phono-target environment. To eliminate this drawback, it is proposed to apply a classifier based on the convolutional neural network, which distributes the selected objects to specific classes.The analysis and experimental studies to evaluate the accuracy of recognition of ground objects by convolutional architectures such as VGG-16, VGG-19, Inception v3, ResNet-50, MobileNet. Training and verification of the recognition quality of architecture data was carried out on an experimentally created data set with a human image on a contrasting background and at different ranges. The results obtained indicate the possibility of using a convolutional neural network in the security system and its ability to work in real time. |
| first_indexed | 2024-04-10T01:20:29Z |
| format | Article |
| id | doaj.art-4de6421322f74651bcc8a26693570ddc |
| institution | Directory Open Access Journal |
| issn | 2309-4923 2414-0481 |
| language | English |
| last_indexed | 2024-04-10T01:20:29Z |
| publishDate | 2020-08-01 |
| publisher | Belarusian National Technical University |
| record_format | Article |
| series | Sistemnyj Analiz i Prikladnaâ Informatika |
| spelling | doaj.art-4de6421322f74651bcc8a26693570ddc2023-03-13T09:47:41ZengBelarusian National Technical UniversitySistemnyj Analiz i Prikladnaâ Informatika2309-49232414-04812020-08-0102465310.21122/2309-4923-2020-2-46-53355Application of convolutional neural networks in the intelligence security system subsystemV. S. Demeshko0A. I. Фёдоров1Военная академия Республики БеларусьВоенная академия Республики БеларусьThe article proposes the structure of an integrated security system for areal facilities. The main subsystems included in its composition are considered. The tasks of the intelligence subsystem for detecting and recognizing ground-based objects of observation in a complex phono-target environment are defined.The task of detecting an object of observation was solved on the basis of a previously proposed algorithm. The disadvantage of this algorithm was the presence of false positives from a flickering complex phono-target environment. To eliminate this drawback, it is proposed to apply a classifier based on the convolutional neural network, which distributes the selected objects to specific classes.The analysis and experimental studies to evaluate the accuracy of recognition of ground objects by convolutional architectures such as VGG-16, VGG-19, Inception v3, ResNet-50, MobileNet. Training and verification of the recognition quality of architecture data was carried out on an experimentally created data set with a human image on a contrasting background and at different ranges. The results obtained indicate the possibility of using a convolutional neural network in the security system and its ability to work in real time.https://sapi.bntu.by/jour/article/view/471подсистема разведкисредства оптико-электронной разведкиизображениеобнаружениераспознаваниеклассификаторсверточная нейронная сетьархитектура |
| spellingShingle | V. S. Demeshko A. I. Фёдоров Application of convolutional neural networks in the intelligence security system subsystem Sistemnyj Analiz i Prikladnaâ Informatika подсистема разведки средства оптико-электронной разведки изображение обнаружение распознавание классификатор сверточная нейронная сеть архитектура |
| title | Application of convolutional neural networks in the intelligence security system subsystem |
| title_full | Application of convolutional neural networks in the intelligence security system subsystem |
| title_fullStr | Application of convolutional neural networks in the intelligence security system subsystem |
| title_full_unstemmed | Application of convolutional neural networks in the intelligence security system subsystem |
| title_short | Application of convolutional neural networks in the intelligence security system subsystem |
| title_sort | application of convolutional neural networks in the intelligence security system subsystem |
| topic | подсистема разведки средства оптико-электронной разведки изображение обнаружение распознавание классификатор сверточная нейронная сеть архитектура |
| url | https://sapi.bntu.by/jour/article/view/471 |
| work_keys_str_mv | AT vsdemeshko applicationofconvolutionalneuralnetworksintheintelligencesecuritysystemsubsystem AT aifëdorov applicationofconvolutionalneuralnetworksintheintelligencesecuritysystemsubsystem |