Experimental results on the UAV123@10fps dataset.
<div><p>Target tracking techniques in the UAV perspective utilize UAV cameras to capture video streams and identify and track specific targets in real-time. Deep learning UAV target tracking methods based on the Siamese family have achieved significant results but still face challenges r...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| مؤلفون آخرون: | , , , , |
| منشور في: |
2025
|
| الموضوعات: | |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| _version_ | 1852023545361596416 |
|---|---|
| author | Yuanhong Dan (20570631) |
| author2 | Jinyan Li (24722) Yu Jin (360487) Yong Ji (296783) Zhihao Wang (473744) Dong Cheng (596943) |
| author2_role | author author author author author |
| author_facet | Yuanhong Dan (20570631) Jinyan Li (24722) Yu Jin (360487) Yong Ji (296783) Zhihao Wang (473744) Dong Cheng (596943) |
| author_role | author |
| dc.creator.none.fl_str_mv | Yuanhong Dan (20570631) Jinyan Li (24722) Yu Jin (360487) Yong Ji (296783) Zhihao Wang (473744) Dong Cheng (596943) |
| dc.date.none.fl_str_mv | 2025-01-16T18:29:32Z |
| dc.identifier.none.fl_str_mv | 10.1371/journal.pone.0314485.g006 |
| dc.relation.none.fl_str_mv | https://figshare.com/articles/figure/Experimental_results_on_the_UAV123_10fps_dataset_/28221393 |
| dc.rights.none.fl_str_mv | CC BY 4.0 info:eu-repo/semantics/openAccess |
| dc.subject.none.fl_str_mv | Neuroscience Sociology Space Science Biological Sciences not elsewhere classified Information Systems not elsewhere classified target deformation motion several popular algorithms physical experimental environment model &# 8217 global attention mechanism capture video streams achieved significant results perform feature fusion global feature interaction feature refinement capability uav tracking datasets track specific targets time processing speed algorithm balances speed feature representation speed compatibility tracking performance small targets view range siamese family quality anchors generate high deep inter correlation operations computational effort comparison experiments |
| dc.title.none.fl_str_mv | Experimental results on the UAV123@10fps dataset. |
| dc.type.none.fl_str_mv | Image Figure info:eu-repo/semantics/publishedVersion image |
| description | <div><p>Target tracking techniques in the UAV perspective utilize UAV cameras to capture video streams and identify and track specific targets in real-time. Deep learning UAV target tracking methods based on the Siamese family have achieved significant results but still face challenges regarding accuracy and speed compatibility. In this study, in order to refine the feature representation and reduce the computational effort to improve the efficiency of the tracker, we perform feature fusion in deep inter-correlation operations and introduce a global attention mechanism to enhance the model’s field of view range and feature refinement capability to improve the tracking performance for small targets. In addition, we design an anchor-free frame-aware feature modulation mechanism to reduce computation and generate high-quality anchors while optimizing the target frame refinement computation to improve the adaptability to target deformation motion. Comparison experiments with several popular algorithms on UAV tracking datasets, such as UAV123@10fps, UAV20L, and DTB70, show that the algorithm balances speed and accuracy. In order to verify the reliability of the algorithm, we built a physical experimental environment on the Jetson Orin Nano platform. We realized a real-time processing speed of 30 frames per second.</p></div> |
| eu_rights_str_mv | openAccess |
| id | Manara_3af6adf76aafe0fe31e08090fe8939af |
| identifier_str_mv | 10.1371/journal.pone.0314485.g006 |
| network_acronym_str | Manara |
| network_name_str | ManaraRepo |
| oai_identifier_str | oai:figshare.com:article/28221393 |
| publishDate | 2025 |
| repository.mail.fl_str_mv | |
| repository.name.fl_str_mv | |
| repository_id_str | |
| rights_invalid_str_mv | CC BY 4.0 |
| spelling | Experimental results on the UAV123@10fps dataset.Yuanhong Dan (20570631)Jinyan Li (24722)Yu Jin (360487)Yong Ji (296783)Zhihao Wang (473744)Dong Cheng (596943)NeuroscienceSociologySpace ScienceBiological Sciences not elsewhere classifiedInformation Systems not elsewhere classifiedtarget deformation motionseveral popular algorithmsphysical experimental environmentmodel &# 8217global attention mechanismcapture video streamsachieved significant resultsperform feature fusionglobal feature interactionfeature refinement capabilityuav tracking datasetstrack specific targetstime processing speedalgorithm balances speedfeature representationspeed compatibilitytracking performancesmall targetsview rangesiamese familyquality anchorsgenerate highdeep intercorrelation operationscomputational effortcomparison experiments<div><p>Target tracking techniques in the UAV perspective utilize UAV cameras to capture video streams and identify and track specific targets in real-time. Deep learning UAV target tracking methods based on the Siamese family have achieved significant results but still face challenges regarding accuracy and speed compatibility. In this study, in order to refine the feature representation and reduce the computational effort to improve the efficiency of the tracker, we perform feature fusion in deep inter-correlation operations and introduce a global attention mechanism to enhance the model’s field of view range and feature refinement capability to improve the tracking performance for small targets. In addition, we design an anchor-free frame-aware feature modulation mechanism to reduce computation and generate high-quality anchors while optimizing the target frame refinement computation to improve the adaptability to target deformation motion. Comparison experiments with several popular algorithms on UAV tracking datasets, such as UAV123@10fps, UAV20L, and DTB70, show that the algorithm balances speed and accuracy. In order to verify the reliability of the algorithm, we built a physical experimental environment on the Jetson Orin Nano platform. We realized a real-time processing speed of 30 frames per second.</p></div>2025-01-16T18:29:32ZImageFigureinfo:eu-repo/semantics/publishedVersionimage10.1371/journal.pone.0314485.g006https://figshare.com/articles/figure/Experimental_results_on_the_UAV123_10fps_dataset_/28221393CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/282213932025-01-16T18:29:32Z |
| spellingShingle | Experimental results on the UAV123@10fps dataset. Yuanhong Dan (20570631) Neuroscience Sociology Space Science Biological Sciences not elsewhere classified Information Systems not elsewhere classified target deformation motion several popular algorithms physical experimental environment model &# 8217 global attention mechanism capture video streams achieved significant results perform feature fusion global feature interaction feature refinement capability uav tracking datasets track specific targets time processing speed algorithm balances speed feature representation speed compatibility tracking performance small targets view range siamese family quality anchors generate high deep inter correlation operations computational effort comparison experiments |
| status_str | publishedVersion |
| title | Experimental results on the UAV123@10fps dataset. |
| title_full | Experimental results on the UAV123@10fps dataset. |
| title_fullStr | Experimental results on the UAV123@10fps dataset. |
| title_full_unstemmed | Experimental results on the UAV123@10fps dataset. |
| title_short | Experimental results on the UAV123@10fps dataset. |
| title_sort | Experimental results on the UAV123@10fps dataset. |
| topic | Neuroscience Sociology Space Science Biological Sciences not elsewhere classified Information Systems not elsewhere classified target deformation motion several popular algorithms physical experimental environment model &# 8217 global attention mechanism capture video streams achieved significant results perform feature fusion global feature interaction feature refinement capability uav tracking datasets track specific targets time processing speed algorithm balances speed feature representation speed compatibility tracking performance small targets view range siamese family quality anchors generate high deep inter correlation operations computational effort comparison experiments |