Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics
Polarimetric decomposition methods are widely used in polarimetric Synthetic Aperture Radar (SAR) data processing for extracting scattering characteristics of targets. However, polarization SAR methods for ship detection still face challenges. The traditional constant false alarm rate (CFAR) detecto...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-02-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/17/4/568 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850081459936690176 |
|---|---|
| author | Canbin Hu Hongyun Chen Xiaokun Sun Fei Ma |
| author_facet | Canbin Hu Hongyun Chen Xiaokun Sun Fei Ma |
| author_sort | Canbin Hu |
| collection | DOAJ |
| description | Polarimetric decomposition methods are widely used in polarimetric Synthetic Aperture Radar (SAR) data processing for extracting scattering characteristics of targets. However, polarization SAR methods for ship detection still face challenges. The traditional constant false alarm rate (CFAR) detectors face sea clutter modeling and parameter estimation problems in ship detection, which is difficult to adapt to the complex background. In addition, neural network-based detection methods mostly rely on single polarimetric-channel scattering information and fail to fully explore the polarization properties and physical scattering laws of ships. To address these issues, this study constructed two novel characteristics: a helix-scattering enhanced (HSE) local component and a multi-scattering intensity difference (MSID) edge component, which are specifically designed to describe ship scattering characteristics. Based on the characteristic differences of different scattering components in ships, this paper designs a context aggregation network enhanced by local and edge component characteristics to fully utilize the scattering information of polarized SAR data. With the powerful feature extraction capability of a convolutional neural network, the proposed method can significantly enhance the distinction between ships and the sea. Further analysis shows that HSE is able to capture structural information about the target, MSID can increase ship–sea separation capability, and an HV channel retains more detailed information. Compared with other decomposition models, the proposed characteristic combination model performs well in complex backgrounds and can distinguish ship from sea more effectively. The experimental results show that the proposed method achieves a detection precision of 93.6% and a recall rate of 91.5% on a fully polarized SAR dataset, which are better than other popular network algorithms, verifying the reasonableness and superiority of the method. |
| format | Article |
| id | doaj-art-2e1d6efa2f0b44e8b65b2e93cb547b3d |
| institution | DOAJ |
| issn | 2072-4292 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj-art-2e1d6efa2f0b44e8b65b2e93cb547b3d2025-08-20T02:44:43ZengMDPI AGRemote Sensing2072-42922025-02-0117456810.3390/rs17040568Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component CharacteristicsCanbin Hu0Hongyun Chen1Xiaokun Sun2Fei Ma3College of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaCollege of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaCollege of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaCollege of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, ChinaPolarimetric decomposition methods are widely used in polarimetric Synthetic Aperture Radar (SAR) data processing for extracting scattering characteristics of targets. However, polarization SAR methods for ship detection still face challenges. The traditional constant false alarm rate (CFAR) detectors face sea clutter modeling and parameter estimation problems in ship detection, which is difficult to adapt to the complex background. In addition, neural network-based detection methods mostly rely on single polarimetric-channel scattering information and fail to fully explore the polarization properties and physical scattering laws of ships. To address these issues, this study constructed two novel characteristics: a helix-scattering enhanced (HSE) local component and a multi-scattering intensity difference (MSID) edge component, which are specifically designed to describe ship scattering characteristics. Based on the characteristic differences of different scattering components in ships, this paper designs a context aggregation network enhanced by local and edge component characteristics to fully utilize the scattering information of polarized SAR data. With the powerful feature extraction capability of a convolutional neural network, the proposed method can significantly enhance the distinction between ships and the sea. Further analysis shows that HSE is able to capture structural information about the target, MSID can increase ship–sea separation capability, and an HV channel retains more detailed information. Compared with other decomposition models, the proposed characteristic combination model performs well in complex backgrounds and can distinguish ship from sea more effectively. The experimental results show that the proposed method achieves a detection precision of 93.6% and a recall rate of 91.5% on a fully polarized SAR dataset, which are better than other popular network algorithms, verifying the reasonableness and superiority of the method.https://www.mdpi.com/2072-4292/17/4/568polarimetric synthetic aperture radar (PolSAR)polarimetric decompositionship detectionstructure characteristicdeep learning |
| spellingShingle | Canbin Hu Hongyun Chen Xiaokun Sun Fei Ma Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics Remote Sensing polarimetric synthetic aperture radar (PolSAR) polarimetric decomposition ship detection structure characteristic deep learning |
| title | Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics |
| title_full | Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics |
| title_fullStr | Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics |
| title_full_unstemmed | Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics |
| title_short | Polarimetric SAR Ship Detection Using Context Aggregation Network Enhanced by Local and Edge Component Characteristics |
| title_sort | polarimetric sar ship detection using context aggregation network enhanced by local and edge component characteristics |
| topic | polarimetric synthetic aperture radar (PolSAR) polarimetric decomposition ship detection structure characteristic deep learning |
| url | https://www.mdpi.com/2072-4292/17/4/568 |
| work_keys_str_mv | AT canbinhu polarimetricsarshipdetectionusingcontextaggregationnetworkenhancedbylocalandedgecomponentcharacteristics AT hongyunchen polarimetricsarshipdetectionusingcontextaggregationnetworkenhancedbylocalandedgecomponentcharacteristics AT xiaokunsun polarimetricsarshipdetectionusingcontextaggregationnetworkenhancedbylocalandedgecomponentcharacteristics AT feima polarimetricsarshipdetectionusingcontextaggregationnetworkenhancedbylocalandedgecomponentcharacteristics |