Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking
Visual object tracking is crucial in both artificial intelligence and computer vision tasks. Hyperspectral videos can provide abundant spectral–spatial–temporal information, which brings new opportunities for improving object tracking performance in complex scenes. In this arti...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11072336/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850077358830125056 |
|---|---|
| author | Na-Na Li Heng-Chao Li Jian-Li Wang Xiong-Fei Geng Jie Pan |
| author_facet | Na-Na Li Heng-Chao Li Jian-Li Wang Xiong-Fei Geng Jie Pan |
| author_sort | Na-Na Li |
| collection | DOAJ |
| description | Visual object tracking is crucial in both artificial intelligence and computer vision tasks. Hyperspectral videos can provide abundant spectral–spatial–temporal information, which brings new opportunities for improving object tracking performance in complex scenes. In this article, we propose a hyperspectral object tracking (HOT) algorithm using low-rank regularized correlation filter (LRCF), which fully leverages the relevance of object features. Specifically, to excavate deep semantic information of object regions, we first apply the band regrouping technology to divide the hyperspectral image into multiple three-channel false-color images and extract deep features via the pretrained network. Furthermore, a multifeature fusion strategy is designed, which integrates deep features with shallow handcrafted features (e.g., histograms of the gradient and color name features) through channel-wise concatenation, leveraging their complementary strengths to enhance the overall feature representation. Finally, we design the LRCF model by introducing low-rank constraint to classify and score search samples for object localization. Experimental results indicate that the LRCF-based tracking method is superior to existing HOT methods, achieving an area under curve score of 0.664 on the HOTC2020 dataset. |
| format | Article |
| id | doaj-art-b9fae949c8a24a239a34f56f91619a6a |
| institution | DOAJ |
| issn | 1939-1404 2151-1535 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj-art-b9fae949c8a24a239a34f56f91619a6a2025-08-20T02:45:49ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118175131752510.1109/JSTARS.2025.358632311072336Low-Rank Regularized Correlation Filter for Hyperspectral Video Object TrackingNa-Na Li0https://orcid.org/0009-0000-8136-6853Heng-Chao Li1https://orcid.org/0000-0002-9735-570XJian-Li Wang2https://orcid.org/0000-0003-4774-4894Xiong-Fei Geng3https://orcid.org/0000-0003-1685-4732Jie Pan4https://orcid.org/0009-0001-2548-4684School of Information Science and Technology, Southwest Jiaotong University, Chengdu, ChinaSchool of Information Science and Technology, Southwest Jiaotong University, Chengdu, ChinaSchool of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu, ChinaChina Waterborne Transport Research Institute, Beijing, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing, ChinaVisual object tracking is crucial in both artificial intelligence and computer vision tasks. Hyperspectral videos can provide abundant spectral–spatial–temporal information, which brings new opportunities for improving object tracking performance in complex scenes. In this article, we propose a hyperspectral object tracking (HOT) algorithm using low-rank regularized correlation filter (LRCF), which fully leverages the relevance of object features. Specifically, to excavate deep semantic information of object regions, we first apply the band regrouping technology to divide the hyperspectral image into multiple three-channel false-color images and extract deep features via the pretrained network. Furthermore, a multifeature fusion strategy is designed, which integrates deep features with shallow handcrafted features (e.g., histograms of the gradient and color name features) through channel-wise concatenation, leveraging their complementary strengths to enhance the overall feature representation. Finally, we design the LRCF model by introducing low-rank constraint to classify and score search samples for object localization. Experimental results indicate that the LRCF-based tracking method is superior to existing HOT methods, achieving an area under curve score of 0.664 on the HOTC2020 dataset.https://ieeexplore.ieee.org/document/11072336/Band regroupinghyperspectral object tracking (HOT)low-rank regularized correlation filter (LRCF)multifeature fusion |
| spellingShingle | Na-Na Li Heng-Chao Li Jian-Li Wang Xiong-Fei Geng Jie Pan Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Band regrouping hyperspectral object tracking (HOT) low-rank regularized correlation filter (LRCF) multifeature fusion |
| title | Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking |
| title_full | Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking |
| title_fullStr | Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking |
| title_full_unstemmed | Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking |
| title_short | Low-Rank Regularized Correlation Filter for Hyperspectral Video Object Tracking |
| title_sort | low rank regularized correlation filter for hyperspectral video object tracking |
| topic | Band regrouping hyperspectral object tracking (HOT) low-rank regularized correlation filter (LRCF) multifeature fusion |
| url | https://ieeexplore.ieee.org/document/11072336/ |
| work_keys_str_mv | AT nanali lowrankregularizedcorrelationfilterforhyperspectralvideoobjecttracking AT hengchaoli lowrankregularizedcorrelationfilterforhyperspectralvideoobjecttracking AT jianliwang lowrankregularizedcorrelationfilterforhyperspectralvideoobjecttracking AT xiongfeigeng lowrankregularizedcorrelationfilterforhyperspectralvideoobjecttracking AT jiepan lowrankregularizedcorrelationfilterforhyperspectralvideoobjecttracking |