An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images
Recently, ship detection in synthetic aperture radar (SAR) images has received extensive attention. Most of the current ship detectors preset dense anchor boxes to achieve spatial alignment with ground-truth (GT) objects. Then, the detector defines the positive and negative samples based on the inte...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-05-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/9/2238 |
| _version_ | 1827671588505387008 |
|---|---|
| author | Hao Shi Zhonghao Fang Yupei Wang Liang Chen |
| author_facet | Hao Shi Zhonghao Fang Yupei Wang Liang Chen |
| author_sort | Hao Shi |
| collection | DOAJ |
| description | Recently, ship detection in synthetic aperture radar (SAR) images has received extensive attention. Most of the current ship detectors preset dense anchor boxes to achieve spatial alignment with ground-truth (GT) objects. Then, the detector defines the positive and negative samples based on the intersection-over-unit (IoU) between the anchors and GT objects. However, this label assignment strategy confuses the learning process of the model to a certain extent and results in suboptimal classification and regression results. In this paper, an adaptive sample assignment (ASA) strategy is proposed to select high-quality positive samples according to the spatial alignment and the knowledge learned from the regression and classification branches. Using our model, the selection of positive and negative samples is more explicit, which achieves better detection performance. A regression guided loss is proposed to further lead the detector to select well-classified and well-regressed anchors as high-quality positive samples by introducing the regression performance as a soft label in the calculation of the classification loss. In order to alleviate false alarms, a feature aggregation enhancement pyramid network (FAEPN) is proposed to enhance multi-scale feature representations and suppress the interference of background noise. Extensive experiments using the SAR ship detection dataset (SSDD) and high-resolution SAR images dataset (HRSID) demonstrate the superiority of our proposed approach. |
| first_indexed | 2024-03-10T03:44:00Z |
| format | Article |
| id | doaj.art-7f26a03808314d0597e7f1f157803573 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T03:44:00Z |
| publishDate | 2022-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-7f26a03808314d0597e7f1f1578035732023-11-23T09:12:39ZengMDPI AGRemote Sensing2072-42922022-05-01149223810.3390/rs14092238An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR ImagesHao Shi0Zhonghao Fang1Yupei Wang2Liang Chen3Radar Research Lab, School of Information and Electronics, Beijing Institute of Technology, Beijing 100081, ChinaRadar Research Lab, School of Information and Electronics, Beijing Institute of Technology, Beijing 100081, ChinaRadar Research Lab, School of Information and Electronics, Beijing Institute of Technology, Beijing 100081, ChinaRadar Research Lab, School of Information and Electronics, Beijing Institute of Technology, Beijing 100081, ChinaRecently, ship detection in synthetic aperture radar (SAR) images has received extensive attention. Most of the current ship detectors preset dense anchor boxes to achieve spatial alignment with ground-truth (GT) objects. Then, the detector defines the positive and negative samples based on the intersection-over-unit (IoU) between the anchors and GT objects. However, this label assignment strategy confuses the learning process of the model to a certain extent and results in suboptimal classification and regression results. In this paper, an adaptive sample assignment (ASA) strategy is proposed to select high-quality positive samples according to the spatial alignment and the knowledge learned from the regression and classification branches. Using our model, the selection of positive and negative samples is more explicit, which achieves better detection performance. A regression guided loss is proposed to further lead the detector to select well-classified and well-regressed anchors as high-quality positive samples by introducing the regression performance as a soft label in the calculation of the classification loss. In order to alleviate false alarms, a feature aggregation enhancement pyramid network (FAEPN) is proposed to enhance multi-scale feature representations and suppress the interference of background noise. Extensive experiments using the SAR ship detection dataset (SSDD) and high-resolution SAR images dataset (HRSID) demonstrate the superiority of our proposed approach.https://www.mdpi.com/2072-4292/14/9/2238synthetic aperture radar (SAR)ship detectionlabel assignmentconvolutional neural network (CNN) |
| spellingShingle | Hao Shi Zhonghao Fang Yupei Wang Liang Chen An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images Remote Sensing synthetic aperture radar (SAR) ship detection label assignment convolutional neural network (CNN) |
| title | An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images |
| title_full | An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images |
| title_fullStr | An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images |
| title_full_unstemmed | An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images |
| title_short | An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images |
| title_sort | adaptive sample assignment strategy based on feature enhancement for ship detection in sar images |
| topic | synthetic aperture radar (SAR) ship detection label assignment convolutional neural network (CNN) |
| url | https://www.mdpi.com/2072-4292/14/9/2238 |
| work_keys_str_mv | AT haoshi anadaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT zhonghaofang anadaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT yupeiwang anadaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT liangchen anadaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT haoshi adaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT zhonghaofang adaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT yupeiwang adaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages AT liangchen adaptivesampleassignmentstrategybasedonfeatureenhancementforshipdetectioninsarimages |