Ship detection and recognition in SAR images with cross-modality domain adaption

被引：0
作者：
Song Y. [1 ]
Li J. [2 ]
Tian T. [1 ]
Tian J. [1 ]
机构：
[1] National Key Laboratory of Science and Technology on Multi-spectral Information Processing, Huazhong University of Science and Technology, Wuhan
来源：
Huazhong Keji Daxue Xuebao (Ziran Kexue Ban)/Journal of Huazhong University of Science and Technology (Natural Science Edition) | 2022年 / 50卷 / 11期
关键词：
domain adaptation; military-civilian ship recognition; ship detection; synthetic aperture radar (SAR) image; transfer learning;
D O I：
10.13245/j.hust.221113
中图分类号：
学科分类号：
摘要：
Aiming at the problems of poor imaging quality and insufficient labeling of synthetic aperture radar (SAR) data，an end-to-end method for ship target detection and recognition in SAR images with cross-modality domain adaption was proposed．First，in view of the problems of low discrimination of target features and high complexity of background information in SAR ship target detection，the fusions of multi-scales feature and local context information，and background suppression were designed to improve the feature expression ability of region of interest-Transformer (RoITransformer)． Then，aiming at the difficulty of labeling attributes in SAR data and lack of labeled samples，a domain adaptation module was designed to implement features alignment at the global and instance level to transfer attributes knowledge from optical data to SAR data across modalities．High resolution SAR ship recognition dataset (HRSSRD) of the nearshore area was built up．Results show that the proposed method could realize an mean average precision of 88.5% on the self-built HRSSRD，and has good performance in ship detection and military-civilian recognition task. © 2022 Huazhong University of Science and Technology. All rights reserved.
引用
收藏
页码：107 / 113
页数：6
相关论文
共 21 条
  • [1] LIN Z, JI K，, LENG X, Squeeze and excitation rank faster R-CNN for ship detection in SAR images[J], IEEE Geoscience and Remote Sensing Letters, 16, 5, pp. 751-755, (2019)
  • [2] CUI Z，, WANG X，, LIU N, Ship detection in large-scale SAR images via spatial shuffle-group enhance attention[J], IEEE Transactions on Geoscience and Remote Sensing, 59, 1, pp. 379-391, (2021)
  • [3] ZHANG T W, ZHANG X L．, High-speed ship detection in SAR images based on a grid convolutional neural network[J], Remote Sensing, 11, 10, (2019)
  • [4] (2017)
  • [5] KANG M, JI K, LENG X, Contextual region-based convolutional neural network with multilayer fusion for SAR ship detection[J], Remote Sensing, 9, 8, (2017)
  • [6] REN S，, HE K, GIRSHICK R, Faster R-CNN：towards real-time object detection with region proposal networks[EB/OL]
  • [7] REDMON J，, FARHADI A．YOLOv3：an incremental improvement
  • [8] JIANG Y，, ZHU X, WANG X, R2CNN：rotational region CNN for orientation robust scene text detection[EB/OL]
  • [9] HAN J, DING J, LI J, Align deep features for oriented object detection[J], IEEE Transactions on Geoscience and Remote Sensing, 60, pp. 1-11, (2022)
  • [10] YANG X，, YAN J，, FENG Z, R3Det： refined single-stage detector with feature refinement for rotating object[J], Proceedings of the AAAI Conference on Artificial Intelligence, 35, 4, pp. 3163-3171, (2021)
123