Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process
With the continuous maturity of unmanned aerial vehicle (UAV) technology, its application is more and more extensive. At the same time, the problem of UAV target tracking has also been widely concerned. Aiming at the problem of low recognition accuracy of small target, a target tracking model of UAV...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10858122/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1823859633564942336 |
|---|---|
| author | Athraa Sabeeh Hasan Allak Jianjun Yi Haider M. Al-Sabbagh Liwei Chen |
| author_facet | Athraa Sabeeh Hasan Allak Jianjun Yi Haider M. Al-Sabbagh Liwei Chen |
| author_sort | Athraa Sabeeh Hasan Allak |
| collection | DOAJ |
| description | With the continuous maturity of unmanned aerial vehicle (UAV) technology, its application is more and more extensive. At the same time, the problem of UAV target tracking has also been widely concerned. Aiming at the problem of low recognition accuracy of small target, a target tracking model of UAV based on siamese neural network (SNN) is studied. Firstly, based on the YOLOv5 recognition model, convolutional attention module and multi-scale feature fusion network are introduced. On the basis of the intersection over union loss, the effective intersection over union loss is proposed to improve the loss function, and an improved YOLOv5 target recognition model is established. Then, a fine-grained classification regression network is proposed, which uses per-pixel classification regression to train the tracker. A target tracking model based on SNN is established by adjusting the results with a fine-tuning module. The results showed that the improved YOLOv5 model combined with the optimized loss function had the highest average accuracy of 47.84% and a frame rate of 28.34fps, which was better than the traditional YOLOv5 model. The recognition accuracy in the fused dataset is 93.12%, with a loss value of less than 0.01, which is superior to YOLOv3, YOLOv4, and traditional YOLOv5 models. The method has strong anti-jamming ability in the acceptable range. The target tracking model based on SNN has the highest tracking accuracy and still has good tracking performance in color image environment, which shows certain feasibility and superiority. To sum up, the model built in this study has good application effects and plays a certain role in promoting the development of the UAV industry. |
| format | Article |
| id | doaj-art-af91ae5f0eef4db0ad1cfe1b3c448064 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-af91ae5f0eef4db0ad1cfe1b3c4480642025-02-11T00:01:09ZengIEEEIEEE Access2169-35362025-01-0113243092432210.1109/ACCESS.2025.353646110858122Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking ProcessAthraa Sabeeh Hasan Allak0https://orcid.org/0009-0005-5304-4124Jianjun Yi1https://orcid.org/0009-0006-0097-4296Haider M. Al-Sabbagh2Liwei Chen3Department of Electromechanical Engineering, University of Technology, Baghdad, IraqSchool of Mechanical and Power Engineering, East China University of Science and Technology, Shanghai, ChinaDepartment of Electrical Engineering, University of Basrah, Basrah, IraqSchool of Mechanical and Power Engineering, East China University of Science and Technology, Shanghai, ChinaWith the continuous maturity of unmanned aerial vehicle (UAV) technology, its application is more and more extensive. At the same time, the problem of UAV target tracking has also been widely concerned. Aiming at the problem of low recognition accuracy of small target, a target tracking model of UAV based on siamese neural network (SNN) is studied. Firstly, based on the YOLOv5 recognition model, convolutional attention module and multi-scale feature fusion network are introduced. On the basis of the intersection over union loss, the effective intersection over union loss is proposed to improve the loss function, and an improved YOLOv5 target recognition model is established. Then, a fine-grained classification regression network is proposed, which uses per-pixel classification regression to train the tracker. A target tracking model based on SNN is established by adjusting the results with a fine-tuning module. The results showed that the improved YOLOv5 model combined with the optimized loss function had the highest average accuracy of 47.84% and a frame rate of 28.34fps, which was better than the traditional YOLOv5 model. The recognition accuracy in the fused dataset is 93.12%, with a loss value of less than 0.01, which is superior to YOLOv3, YOLOv4, and traditional YOLOv5 models. The method has strong anti-jamming ability in the acceptable range. The target tracking model based on SNN has the highest tracking accuracy and still has good tracking performance in color image environment, which shows certain feasibility and superiority. To sum up, the model built in this study has good application effects and plays a certain role in promoting the development of the UAV industry.https://ieeexplore.ieee.org/document/10858122/Siamese neural networktarget trackingtarget recognitionunmanned aerial vehicledeep learning |
| spellingShingle | Athraa Sabeeh Hasan Allak Jianjun Yi Haider M. Al-Sabbagh Liwei Chen Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process IEEE Access Siamese neural network target tracking target recognition unmanned aerial vehicle deep learning |
| title | Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process |
| title_full | Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process |
| title_fullStr | Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process |
| title_full_unstemmed | Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process |
| title_short | Siamese Neural Networks in Unmanned Aerial Vehicle Target Tracking Process |
| title_sort | siamese neural networks in unmanned aerial vehicle target tracking process |
| topic | Siamese neural network target tracking target recognition unmanned aerial vehicle deep learning |
| url | https://ieeexplore.ieee.org/document/10858122/ |
| work_keys_str_mv | AT athraasabeehhasanallak siameseneuralnetworksinunmannedaerialvehicletargettrackingprocess AT jianjunyi siameseneuralnetworksinunmannedaerialvehicletargettrackingprocess AT haidermalsabbagh siameseneuralnetworksinunmannedaerialvehicletargettrackingprocess AT liweichen siameseneuralnetworksinunmannedaerialvehicletargettrackingprocess |