SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images
Arbitrary-oriented object detection in remote sensing images has witnessed significant progress in recent years. Numerous excellent detection models perform promising results, however, there are two main tough challenges hinder their performances. On the one hand, current label assignment strategies...
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/10848126/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849702279175733248 |
|---|---|
| author | Yan Dong Minghong Wei Guangshuai Gao Chunlei Li Zhoufeng Liu |
| author_facet | Yan Dong Minghong Wei Guangshuai Gao Chunlei Li Zhoufeng Liu |
| author_sort | Yan Dong |
| collection | DOAJ |
| description | Arbitrary-oriented object detection in remote sensing images has witnessed significant progress in recent years. Numerous excellent detection models perform promising results, however, there are two main tough challenges hinder their performances. On the one hand, current label assignment strategies suffer from an imbalance between positive and negative samples, particularly for large aspect ratio and small-scale objects, leading to the Insufficient High-quality Samples. On the other hand, fixed convolution kernels and coarse sampling positions are not well suited for adapting to rotating objects in complex remote sensing scenes, resulting in Feature Misalignment. To alleviate the above issues, in this article, a novel SARFA-Net is proposed, incorporating a Shape-Aware Label Assignment (SALA) strategy and Refined Feature Alignment module (RFAM). Specifically, SALA is proposed to mitigate the problem of insufficient sampling for extremely shaped objects, the core of which is the Shape-Aware Sampling module, to meticulously select more high-quality positive samples within elliptical regions. To further enhance SALA at extremely limited scales and large aspect ratios, a Threshold Compensation Module is designed, which further utilizes the shape characteristics of the objects. Furthermore, RFAM is developed to adaptively align features by adjusting the sampling positions of the convolution kernels based on the refined anchors. Extensive experiments conducted on five large-scale datasets, DIOR-R, DOTA-v1.0, HRSC2016, FAIR1M-v1.0, and UCAS-AOD achieved mAPs of 68.90%, 80.09%, 90.40%, 46.34%, and 90.01%, respectively, demonstrating the effectiveness of the proposed approach and the superiority compared with state-of-the-arts. Compared with the baseline <inline-formula><tex-math notation="LaTeX">${\text{S}^{\text{2}}}\text{A-Net}$</tex-math></inline-formula>, we have improved by 1.30, 1.57, 0.23, 5.92, and 0.37 points, respectively, without additional data augmentation. |
| format | Article |
| id | doaj-art-ef9c11359f954308bdc6d2bdf02c3fc9 |
| 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-ef9c11359f954308bdc6d2bdf02c3fc92025-08-20T03:17:43ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01188865888110.1109/JSTARS.2025.353203910848126SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing ImagesYan Dong0https://orcid.org/0000-0002-5652-2102Minghong Wei1https://orcid.org/0009-0002-6665-7577Guangshuai Gao2https://orcid.org/0000-0002-5050-2311Chunlei Li3https://orcid.org/0000-0001-6543-1838Zhoufeng Liu4School of Information and Communication Engineering, Zhongyuan University of Technology, Zhengzhou, ChinaSchool of Information and Communication Engineering, Zhongyuan University of Technology, Zhengzhou, ChinaSchool of Information and Communication Engineering, Zhongyuan University of Technology, Zhengzhou, ChinaSchool of Information and Communication Engineering, Zhongyuan University of Technology, Zhengzhou, ChinaSchool of Information and Communication Engineering, Zhongyuan University of Technology, Zhengzhou, ChinaArbitrary-oriented object detection in remote sensing images has witnessed significant progress in recent years. Numerous excellent detection models perform promising results, however, there are two main tough challenges hinder their performances. On the one hand, current label assignment strategies suffer from an imbalance between positive and negative samples, particularly for large aspect ratio and small-scale objects, leading to the Insufficient High-quality Samples. On the other hand, fixed convolution kernels and coarse sampling positions are not well suited for adapting to rotating objects in complex remote sensing scenes, resulting in Feature Misalignment. To alleviate the above issues, in this article, a novel SARFA-Net is proposed, incorporating a Shape-Aware Label Assignment (SALA) strategy and Refined Feature Alignment module (RFAM). Specifically, SALA is proposed to mitigate the problem of insufficient sampling for extremely shaped objects, the core of which is the Shape-Aware Sampling module, to meticulously select more high-quality positive samples within elliptical regions. To further enhance SALA at extremely limited scales and large aspect ratios, a Threshold Compensation Module is designed, which further utilizes the shape characteristics of the objects. Furthermore, RFAM is developed to adaptively align features by adjusting the sampling positions of the convolution kernels based on the refined anchors. Extensive experiments conducted on five large-scale datasets, DIOR-R, DOTA-v1.0, HRSC2016, FAIR1M-v1.0, and UCAS-AOD achieved mAPs of 68.90%, 80.09%, 90.40%, 46.34%, and 90.01%, respectively, demonstrating the effectiveness of the proposed approach and the superiority compared with state-of-the-arts. Compared with the baseline <inline-formula><tex-math notation="LaTeX">${\text{S}^{\text{2}}}\text{A-Net}$</tex-math></inline-formula>, we have improved by 1.30, 1.57, 0.23, 5.92, and 0.37 points, respectively, without additional data augmentation.https://ieeexplore.ieee.org/document/10848126/Arbitrary-oriented object detection (AOOD)feature misalignment (FM)insufficient high-quality sampleslabel assignmentremote sensing images (RSI) |
| spellingShingle | Yan Dong Minghong Wei Guangshuai Gao Chunlei Li Zhoufeng Liu SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Arbitrary-oriented object detection (AOOD) feature misalignment (FM) insufficient high-quality samples label assignment remote sensing images (RSI) |
| title | SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images |
| title_full | SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images |
| title_fullStr | SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images |
| title_full_unstemmed | SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images |
| title_short | SARFA-Net: Shape-Aware Label Assignment and Refined Feature Alignment for Arbitrary-Oriented Object Detection in Remote Sensing Images |
| title_sort | sarfa net shape aware label assignment and refined feature alignment for arbitrary oriented object detection in remote sensing images |
| topic | Arbitrary-oriented object detection (AOOD) feature misalignment (FM) insufficient high-quality samples label assignment remote sensing images (RSI) |
| url | https://ieeexplore.ieee.org/document/10848126/ |
| work_keys_str_mv | AT yandong sarfanetshapeawarelabelassignmentandrefinedfeaturealignmentforarbitraryorientedobjectdetectioninremotesensingimages AT minghongwei sarfanetshapeawarelabelassignmentandrefinedfeaturealignmentforarbitraryorientedobjectdetectioninremotesensingimages AT guangshuaigao sarfanetshapeawarelabelassignmentandrefinedfeaturealignmentforarbitraryorientedobjectdetectioninremotesensingimages AT chunleili sarfanetshapeawarelabelassignmentandrefinedfeaturealignmentforarbitraryorientedobjectdetectioninremotesensingimages AT zhoufengliu sarfanetshapeawarelabelassignmentandrefinedfeaturealignmentforarbitraryorientedobjectdetectioninremotesensingimages |