NEURAL NETWORK RECOGNITION OF MILITARY VEHICLES IN AERIAL IMAGES USING A MODIFIED YOLOV8 ARCHITECTURE AND FREQUENCY-SPATIAL FEATURES
DOI:
https://doi.org/10.32782/mathematical-modelling/2026-9-1-14Keywords:
neural networks, deep learning, YOLO, aerial images, military vehicles, object recognitionAbstract
This paper presents a modified YOLOv8-based object detection architecture for automated recognition of military transport vehicles in high-resolution aerospace imagery. The proposed approach aims to improve detection accuracy under conditions of complex background, low object contrast, and arbitrary spatial orientation. The architecture integrates a frequency–spatial feature enhancement module (Freq-SpaFEM), a modified bidirectional feature pyramid network (BiFPN), and oriented bounding boxes with rotation parameter estimation. The training process was conducted over 320 epochs, ensuring stable convergence and high generalization performance on validation data. Experimental results demonstrate that the proposed model achieves superior detection performance, reaching mAP@50 of 98.6 %, Precision of 98.7 %, and Recall of 95.7 %. Comparative analysis with baseline models, including YOLOv5, YOLOv9, YOLOv11, and the standard YOLOv8, confirms significant improvements in detection accuracy, reductions in false positives and false negatives, and enhanced robustness to challenging scene conditions. The scientific novelty of the study lies in the development of an integrated approach that combines frequency-domain and spatial feature analysis within a unified deep learning framework, along with the implementation of rotation-aware feature alignment and oriented bounding- box parameterization for aerospace object detection. It enables more accurate representation of object geometry and improves detection performance for partially occluded and arbitrarily oriented targets. Analysis of error matrices confirmed the high ability of the model to distinguish objects and the background environment, which is manifested in a high level of true positive results and a minimum level of false positives. This indicates the effectiveness of using the proposed modules to increase the informativeness of features and localization accuracy. A software application based on deep learning libraries and computer vision tools was developed to validate the proposed method in practice, providing realtime visualization and analysis capabilities. The obtained results confirm the effectiveness of the proposed method for aerospace image analysis and highlight its potential for practical deployment in monitoring, surveillance, and geospatial intelligence systems.
References
Robinson K., McKenna B., Rooney D. The relationship of risk to rules, values, virtues, and moral complexity: What we can learn from the moral struggles of military leaders. J. Bus. Ethics. 2022. Vol. 179. P. 749–766. https://doi.org/10.1007/s10551-021-04842-4
Peng H., Zhang Y., Yang S., Song B. Battlefield image situational awareness application based on deep learning. IEEE Intell. Syst. 2019. Vol. 35. P. 36–43. https://doi.org/10.1109/MIS.2019.2942701
Mei C., Fan Z., Zhu Q., Yang P., Hou Z., Jin H. A Novel scene matching navigation system for UAVs based on vision/inertial fusion. IEEE Sens. J. 2023. Vol. 23. P. 6192–6203. https://doi.org/10.1109/JSEN.2023.3242045
Zhou L., Leng S., Liu Q., Wang Q. Intelligent UAV swarm cooperation for multiple targets tracking. IEEE Internet Things J. 2021. Vol. 9. P. 743–754. https://doi.org/10.1109/JIOT.2021.3106339
Meng Q., Zhang L., Sun Z., Meng F., Wang L., Sun Y. Characterizing spatial and temporal trends of surface urban heat island effect in an urban main built-up area: A 12-year case study in Beijing, China. Remote Sens. Environ. 2018. Vol. 204. P. 826–837. https://doi.org/10.1016/j.rse.2017.10.027
Scuro V. Open-Source Intelligence (OSINT) for Researchers and Practitioners. 2025. https://doi.org/10.1007/978-3-032-02014-7_2
Carion N., Massa F., Synnaeve G. et al. End-to-end object detection with transformers. In Proceedings of the European Conference on Computer Vision (ECCV). Glasgow, UK, 23–28 August 2020. Springer, Cham. P. 213–229. https://doi.org/10.1007/978-3-030-58452-8_13
Wang W. et al. Pyramid vision transformer: A versatile backbone for dense prediction without convolutions. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). Virtual, 11–17 October 2021. P. 568–578. https://doi.org/10.1109/ICCV48922.2021.00061
Sumari A. D. W. et al. Automatic target recognition and identification for military ground-toair observation tasks using support vector machine and information fusion. In Proceedings of the 2022 International Conference on ICT for Smart Society (ICISS). Bandung, Indonesia, 10–11 August 2022. P. 1–8. https://doi.org/10.1109/ICISS55894.2022.9915201
Su H. et al. Object detection and instance segmentation in remote sensing imagery based on precise mask R-CNN. In Proceedings of the IGARSS 2019 – 2019 IEEE International Geoscience and Remote Sensing Symposium. Yokohama, Japan, 28 July–2 August 2019. P. 1454–1457. https://doi.org/10.1109/IGARSS.2019.8898160
Guo Z., Mu X., Chang C. et al. An Enhanced Framework for Small Object Detection with MiddleOrder Interaction and Adaptive Cross-Scale Aggregation. Engineering Applications of Artificial Intelligence. 2025. Vol. 130. https://doi.org/10.1016/j.engappai.2024.107736
Zhao X., Xia Y., Zhang W., Zheng C., Zhang Z. YOLO-ViT-Based Method for Unmanned Aerial Vehicle Infrared Vehicle Target Detection. Remote Sens. 2023. Vol. 15. 3778. https://doi.org/10.3390/rs15153778
Alrayes F. S., Ahmad N., Alshuhail A. et al. Convolutional Transform Learning-Based Fusion for Scale-Invariant Long-Term Target Detection in UAVs. Scientific Reports. 2025. Vol. 15. https://doi.org/10.1038/s41598-024-51123-x
Chutke S., Kumar V. V., Saravanakumar M. AI-Enhanced Image Processing for Target Detection and Threat Recognition in Defence Applications. In Proc. IEEE Conf. on Smart Technologies and Virtual Environments. 2025.
Wang J., Wu Y., Liu B. et al. Fourier aids CNN and transformer for semantic segmentation of remote sensing images. Multimedia Systems. 2025. Vol. 31. 451. https://doi.org/10.1007/s00530-025-02029-4
Tan M., Pang R., Le Q. EfficientDet: Scalable and Efficient Object Detection. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2020. P. 10778-10787. https://doi.org/10.1109/CVPR42600.2020.01079
Afifah V., Erniwati S. YOLOv8 for Object Detection: A Comprehensive Review of Advances, Techniques, and Applications. International Journal of Advanced Computing and Informatics. 2025. Vol. 2. P. 53–61. https://doi.org/10.71129/ijaci.v2i1.pp53-61
