Derin Öğrenme Mimarileriyle Gerçek Zamanlı Hava Görüntülerinde Çoklu Nesne Tespiti ve Takibi

Abstract

Gerçek zamanlı sistemlerde çoklu nesne tespiti ve takibi, insansız hava araçları (İHA) ve uydu sistemleri gibi platformlardan elde edilen hava görüntülerinin analizinde karşılaşılan başlıca zorluklardan biridir. Geleneksel takip algoritmaları, özellikle karmaşık ve dinamik ortamlarda yetersiz kalmakta, bu da daha güçlü ve esnek yöntemlerin geliştirilmesini gerekli kılmaktadır. Bu çalışma, yüksek çıkarım oranı ve yeterli doğruluk hedefleri doğrultusunda derin öğrenme tabanlı bir çözüm önermektedir. Çalışma kapsamında, tek aşamalı nesne tespit mimarisi olan YOLOv11 modeli, tespit edilen nesnelerin takibi için ByteTrack algoritmasıyla entegre edilmiştir. Bu yaklaşım, nesne tespiti konusunda yüksek doğruluğuyla bilinen iki aşamalı Faster R-CNN mimarisi (MobileNetV3-large FPN omurgası) ile karşılaştırmalı olarak değerlendirilmiştir. Modellerin eğitimi, farklı boyutlardaki nesneleri içeren zorlu VisDrone ve DOTA veri setlerinin birleştirilmesiyle oluşturulan kapsamlı bir veri kümesi üzerinde gerçekleştirilmiştir. Sonuçlar, YOLOv11 modelinin yüksek hızlı çıkarım yeteneği sayesinde gerçek zamanlı uygulamalar için daha uygun olduğunu, buna karşın Faster R-CNN modelinin daha yüksek doğrulukta tespitler sağladığını ancak daha fazla hesaplama maliyeti gerektirdiğini göstermektedir. Takip aşamasında kullanılan ByteTrack algoritması, tespit edilen nesneleri başarılı bir şekilde tanımlayarak takip doğruluğunu artırmıştır. Bu bağlamda, her iki modelin avantajları ve sınırlılıkları değerlendirilmiş ve model seçiminin hedeflenen uygulamanın gereksinimlerine göre yapılması gerektiği sonucuna varılmıştır.

Keywords

• Çoklu Nesne Takibi
• YOLOv11
• Faster R-CNN
• ByteTrack
• Havasal Görüntüler

Multiple Object Detection and Tracking in Real-Time Aerial Imagery with Deep Learning Architectures

Abstract

Real-time multi-object detection and tracking is one of the main challenges in analysing aerial imagery from platforms such as unmanned aerial vehicles (UAVs) and satellite systems. Traditional tracking algorithms are inadequate especially in complex and dynamic environments, which necessitates the development of more powerful and flexible methods. This study proposes a deep learning based solution in line with the objectives of high extraction rate and sufficient accuracy. Within the scope of the study, the YOLOv11 model, which is a single-stage object detection architecture, is integrated with the ByteTrack algorithm for tracking the detected objects. This approach is evaluated in comparison with the two-stage Faster R-CNN architecture (MobileNetV3-large FPN backbone), which is known for its high accuracy in object detection. The training of the models was performed on a comprehensive dataset created by combining the challenging VisDrone and DOTA datasets containing objects of different sizes. The results show that the YOLOv11 model is more suitable for real-time applications due to its high speed inference capability, while the Faster R-CNN model provides more accurate detections despite its higher computational cost. The ByteTrack algorithm used in the tracking phase has increased the tracking accuracy by successfully identifying the detected objects. In this context, the advantages and limitations of both models are evaluated and it is concluded that the choice of model should be made according to the requirements of the targeted application.

Keywords

• Multi-Object Tracking
• YOLOv11
• Faster R-CNN
• ByteTrack
• Aerial Images

References

1. [1] Simonyan, Karen & Zisserman, Andrew. ( 2014 ). Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv 1409.1556.].
2. [2] Inik, O., İnik, Ö., Öztaş, T., Demir, Y., & Yüksel, A. (2023). Prediction of soil organic matter with deep learning. Arabian Journal for Science and Engineering, 48(8), 10227-10247.
3. [3] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25.
4. [4] İnik, Ö., Uyar, K., & Ülker, E. (2018). Gender classification with a novel convolutional neural network (CNN) model and comparison with other machine learning and deep learning CNN models. Journal Of Industrial Engineering Research, 4(4), 57-63.
5. [5] Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).
6. [6] İnik, Ö., & Turan, B. (2018). Classification of animals with different deep learning models. Journal of New Results in Science, 7(1), 9-16.
7. [7] Liu, Z., Chen, Y., & Gao, Y. (2024). Rotating-YOLO: A novel YOLO model for remote sensing rotating object detection. Image and Vision Computing, 105397.
8. [8] Wang, T., Ma, Z., Yang, T., & Zou, S. (2023). PETNet: A YOLO-based prior enhanced transformer network for aerial image detection. Neurocomputing, 547, 126384.

9. [9] Xue, C., Xia, Y., Wu, M., Chen, Z., Cheng, F., & Yun, L. (2024). EL-YOLO: An efficient and lightweight low-altitude aerial objects detector for onboard applications. Expert Systems with Applications, 256, 124848.
10. [10] Battish, N., Kaur, D., Chugh, M., & Poddar, S. (2024). SDMNet: spatially dilated multi-scale network for object detection for drone aerial imagery. Image and Vision Computing, 150, 105232.
11. [11] Luo, W., & Yuan, S. (2025). Enhanced YOLOv8 for small-object detection in multiscale UAV imagery: Innovations in detection accuracy and efficiency. Digital Signal Processing, 158, 104964.
12. [12] Qu, Z., Liu, H., Kong, W., Gu, J., Wang, C., Deng, L., ... & Lin, F. (2025). LP-YOLO: An improved lightweight pedestrian detection algorithm based on YOLOv11. Digital Signal Processing, 105343.
13. [13] Wang, D., Tan, J., Wang, H., Kong, L., Zhang, C., Pan, D., ... & Liu, J. (2025). SDS-YOLO: An improved vibratory position detection algorithm based on YOLOv11. Measurement, 244, 116518.
14. [14] Ramos, A., Moraes, F., & Martins, R. (2023). Fire and smoke detection in ground and aerial images using YOLObased architectures. Fire Technology, 59, 2091–2113.
15. [15] Li, R., Yu, J., Li, F., Yang, R., Wang, Y., & Peng, Z. (2023). Automatic bridge crack detection using Unmanned aerial vehicle and Faster R-CNN. Construction and Building Materials, 362, 129659.
16. [16] Cui, J., Zhang, X., Zhang, J., Han, Y., Ai, H., Dong, C., & Liu, H. (2024). Weed identification in soybean seedling stage based on UAV images and Faster R-CNN. Computers and Electronics in Agriculture, 227, 109533.
17. [17] Li, L., Hassan, M. A., Yang, S., Jing, F., Yang, M., Rasheed, A., ... & Xiao, Y. (2022). Development of imagebased wheat spike counter through a Faster R-CNN algorithm and application for genetic studies. The Crop Journal, 10(5), 1303-1311.
18. [18] Ma, Y., Chai, L., Jin, L., & Yan, J. (2024). Hierarchical alignment network for domain adaptive object detection in aerial images. ISPRS Journal of Photogrammetry and Remote Sensing, 208, 39-52.
19. [19] Luo, W., & Yuan, S. (2025). Enhanced YOLOv8 for small-object detection in multiscale UAV imagery: Innovations in detection accuracy and efficiency. Digital Signal Processing, 158, 104964.
20. [20] Gu, Q., Han, Z., Kong, S., Huang, H., Li, Y., Fan, Q., & Wu, R. (2025). DCYOLO: Dual negative weighting label assignment and cross-layer decouple head for YOLO in remote sensing images. Expert Systems with Applications, 281, 127595.
21. [21] Chen, Z., Wang, H., Wu, X., Wang, J., Lin, X., Wang, C., ... & Li, D. (2024). Object detection in aerial images using DOTA dataset: A survey. International Journal of Applied Earth Observation and Geoinformation, 134, 104208.
22. [22] Nguyen, K., Huynh, N. T., Le, D. T., Huynh, D. T., Bui, T. T. T., Dinh, T., ... & Nguyen, T. V. (2025). A comprehensive review of few-shot object detection on aerial imagery. Computer Science Review, 57, 100760.
23. [23] Han, L., Li, N., Zhong, Z., Niu, D., & Gao, B. (2025). Adaptive scale matching for remote sensing object detection based on aerial images. Image and Vision Computing, 157, 105482.
24. [24] Jobaer, S., Tang, X. S., & Zhang, Y. (2025). A deep neural network for small object detection in complex environments with unmanned aerial vehicle imagery. Engineering Applications of Artificial Intelligence, 148, 110466.
25. [25] Xue, C., Xia, Y., Wu, M., Chen, Z., Cheng, F., & Yun, L. (2024). EL-YOLO: An efficient and lightweight low-altitude aerial objects detector for onboard applications. Expert Systems with Applications, 256, 124848.
26. [26] Li, Y., Zhang, W., Lv, S., Yu, J., Ge, D., Guo, J., & Li, L. (2025). YOLOv11-CAFM model in ground penetrating radar image for pavement distress detection and optimization study. Construction and Building Materials, 485, 141907.
27. [27] Tetila, E. C., Junior, G. W., Higa, G. T. H., da Costa, A. B., Amorim, W. P., Pistori, H., & Barbedo, J. G. A. (2025). Deep learning models for detection and recognition of weed species in corn crop. Crop Protection, 107237.
28. [28] He, L. H., Zhou, Y. Z., Liu, L., Zhang, Y. Q., & Ma, J. H. (2025). Research on the directional bounding box algorithm of YOLO11 in tailings pond identification. Measurement, 117674.
29. [29] Zhang, Y., Sun, P., Jiang, Y., Yu, D., Weng, F., Yuan, Z., ... & Wang, X. (2022, October). Bytetrack: Multi-object tracking by associating every detection box. In European conference on computer vision (pp. 1-21). Cham: Springer Nature Switzerland.