An Adaptive Sample Assignment Strategy Based on Feature Enhancement for Ship Detection in SAR Images
Open Access
- 7 May 2022
- journal article
- research article
- Published by MDPI AG in Remote Sensing
- Vol. 14 (9) , 2238
Abstract
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.Keywords
Funding Information
- National Natural Science Foundation of China (62101041, T2012122, Z141101001514005)
- Chang Jiang Scholars Program (62101041, T2012122, Z141101001514005)
- Hundred Leading Talent Project of Beijing Science and Technology (62101041, T2012122, Z141101001514005)
This publication has 41 references indexed in Scilit:
- Saliency-Guided Single Shot Multibox Detector for Target Detection in SAR ImagesIEEE Transactions on Geoscience and Remote Sensing, 2019
- Dense Attention Pyramid Networks for Multi-Scale Ship Detection in SAR ImagesIEEE Transactions on Geoscience and Remote Sensing, 2019
- Ship Detection Based on YOLOv2 for SAR ImageryRemote Sensing, 2019
- A SAR Dataset of Ship Detection for Deep Learning under Complex BackgroundsRemote Sensing, 2019
- Demystifying the Capability of Sublook Correlation Techniques for Vessel Detection in SAR ImageryIEEE Transactions on Geoscience and Remote Sensing, 2018
- Superpixel-Level CFAR Detectors for Ship Detection in SAR ImageryIEEE Geoscience and Remote Sensing Letters, 2018
- Focal Loss for Dense Object Detection2017
- SSD: Single Shot MultiBox Detector2016
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal NetworksIEEE Transactions on Pattern Analysis and Machine Intelligence, 2016
- The Pascal Visual Object Classes (VOC) ChallengeInternational Journal of Computer Vision, 2009