基于可变形的多尺度自注意力特征融合SAR影像舰船识别

周慧, 朱虹, 陈澎

doi:10.16411/j.cnki.issn1006-7736.2024.04.012

大连海事大学学报 >

2024 , Vol. 50 >Issue 4: 110 - 118

DOI: https://doi.org/10.16411/j.cnki.issn1006-7736.2024.04.012

基于可变形的多尺度自注意力特征融合SAR影像舰船识别

展开

(1.大连东软信息学院计算机与软件学院，辽宁大连 116023;2.大连海事大学航海学院，辽宁大连 116026)

周慧（1983 — ），女，硕士，教授，研究方向：遥感图像处理，E—mail：zhouhui@neusoft.edu.cn。陈澎*（1982 — ），男，博士，副教授，研究方向：遥感与地理信息科学，E-mail：chenpeng@dlmu.edu.cn。

网络出版日期: 2024-06-05

基金资助

辽宁省教育厅基本科研项目面上项目（LJKMZ20222006）

收起

Deformable-based multi-scale self-attentive feature fusion for SAR image ship recognition

Expand

(1.School of Computer and Software, Dalian Neusoft Information University, Dalian 116023, China;2.Navigation College, Dalian Maritime University, Dalian 116026, China)

Online published: 2024-06-05

Fold

摘要

SAR图像中不同类别船舶的目标特征区分不明显，当船舶类别较多时会出现识别准确率下降的问题。为更好地提取类别特征，本文提出一种识别模型DCN-MSFF-TR，借鉴Transformer encoder-decoder思想，在主干网络中加入可变形卷积模块（DCN），同时，将经过Transformer多尺度自注意力处理的特征层按照特征金字塔的方式在模型合适的位置进行特征融合，每一层不仅能够利用自身的信息，还能够综合利用其他层的特征。在公开数据集Open SARShip-3-Complex三分类数据集和Open SARShip-6-Complex六分类数据集的验证结果表明，平均识别精确率分别达到78.1%和66.7%，说明本文方法相对其他识别模型能更有效识别出SAR图像中的舰船类别。

关键词： SAR图像; 船舶识别; Transformer架构; 可变卷积; 特征融合

本文引用格式

周慧, 朱虹, 陈澎 . 基于可变形的多尺度自注意力特征融合SAR影像舰船识别[J]. 大连海事大学学报, 2024 , 50(4) : 110 -118 . DOI: 10.16411/j.cnki.issn1006-7736.2024.04.012

Abstract

The feature differentiation of different categories of ship targets in SAR images is not clear, and the recognition accuracy may decrease when there are many ship categories. To better extract category features, this paper proposed a recognition model DCN-MSF-TR, which drawed on the idea of Transformer encoder-decoder and added a deformable convolutional module (DCN) to the backbone network. At the same time, the feature layers processed by Transformer multi-scale self attention were fused at appropriate positions in the model in a feature pyramid manner, and each layer can not only utilize its own information, but also comprehensively utilize the features of other layers. The validation results on the Open SARShip-3-Complex three class dataset and Open SARShip-6-Complex six class dataset show that the average recognition accuracy reaches 78.1% and 66.7%, respectively, which show that the proposed method can more effectively identify ship categories in SAR images compared to other recognition models.

Key words： synthetic apeture radar(SAR) images; ship recognition; transformer architecture; deformable convolution; feature fusion

参考文献

[1]XIAO X W, ZHOU Z Q, WANG B, et al. Ship detection under complex backgrounds based on accurate rotated anchor boxes from paired semantic segmentation[J]. Remote Sensing,2019, 11:2506-2528.

[2]刘涛,杨子渊.极化SAR图像舰船目标检测研究综述[J].雷达学报, 2021, 10(1):1-19.

LIU T, YANG Z Y. A review of ship target detection in polarimetric SAR images[J]. Journal of Radars, 2021, 10(1), 1-19. (in Chinese)

[3]LI Y, DU L, WEI D. Multiscale CNN based on component analysis for SAR ATR[J]. IEEE Transactions on Geoscience and Remote Sensing，2021, 60:1-24.

[4]FU Q, LUO K M, SONG Y, et al. Study of sea fog environment polarization transmission characteristics[J]. Applied Sciences, 2022, 12(17)： 8892.

[5]CHEN P, LI Y, ZHOU H, et al. Detection of small ship objects using anchor boxes cluster and feature pyramid network model for SAR imagery[J]. Journal of Marine Science and Engineering, 2020, 8(2) :112-125.

[6]刘婷,罗佩琪,范云生.基于SSD的海面小目标检测综述[J].大连海事大学学报,2022,48(4):65-75.

LIU T, LUO P Q, FAN Y S. A review of SSD based small object detection on the sea surface[J]. Journal of Dalian Maritime University, 2022, 48(4):65-75. (in Chinese)

[7]GRAZIANO M D, RENGA A, MOCCIA A. Integration of automatic identification system (AIS) data and single-channel synthetic aperture radar (SAR) images by SAR based ship velocity estimation for maritime situational awareness[J]. Remote Sensing，2019, 11(19):2196.

[8]丘锐聪,周海峰,陈颖,等.融合可切换空洞卷积的上下文信息增强船舶目标检测算法[J]. 大连海事大学学报, 2023, 49(4):116--125.

QIU R C, ZHOU H F, CHEN Y, et al. Ship target detection algorithm based on context augmentation information with switchable dilated convolution[J]. Journal of Dalian Maritime University, 2023, 49(4): 116-125. (in Chinese)

[9]LIU W B, WANG Z D, LIU X H, et al. A survey of deep neural network architectures and their applications[J].Neurocomputing, 2017, 234:11-26.

[10]ZENG L, ZHU Q T, LU D W, et al. Dual-polarized SAR ship grained classification based on CNN with hybrid channel feature loss[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19:1-5.

[11]WANG C, SHI J, ZHOU Y Y, et al. Semi-supervised learning-based SAR ATR via self-consistent augmentation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(6):4862-4873.

[12]HE J L, WANG Y H, LIU H W. Ship classification in medium-resolution SAR images via densely connected triplet CNNs integrating Fisher discrimination regularized metric learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59(4):3022-3039.

[13]HUANG G Q, LIU X, HUI J, et al. A novel group squeeze excitation sparsely connected convolutional networks for SAR target classification[J]. International Journal of Remote Sensing, 2019, 40(11):4346-4360.

[14]DONG Y B, ZHANG H, WANG C, et al. Fine-grained ship classification based on deep residual learning for high-resolution SAR images[J]. Remote Sensing Letter, 2019, 10(11):1095-1104.

[15]LI J W, QU C W, PENG S J. Ship classification for unbalanced SAR dataset based on convolutional neural network[J]. Journal of Applied Remote Sensing, 2018, 12(3):035010.

[16]HOU X Y, AO W, SONG Q, et al. FUSAR-Ship:building a high-resolution SAR-AIS matchup dataset of Gaofen-3 for ship detection and recognition[J]. Science China Information Sciences, 2020, 63:1-9.

[17]CHEN Y T, WANG J L, ZHANG Y Y, et al. P2RNet:fast maritime object detection from key points to region proposals in large-scale remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024,17: 9294-9308.

[18]田永林, 王雨桐, 王建功, 等. 视觉Transformer 研究的关键问题: 现状及展望[J]. 自动化学报, 2022, 48(4):957-979.

TIAN Y L, WANG Y T, WANG J G, et al. Key problems and progress of vision transformers: the state of the art and prospects[J]. Acta Automatica Sinica, 2022, 48(4): 957-979. (in Chinese)

[19]SUN Z Q, MENG C N, CHENG J R, et al. Amulti-scale feature pyramid network for detection and instance segmentation of marine ships in SAR images[J]. Remote Sensing, 2022, 14: 6312.

[20]CHEN P, ZHOU H, LI Y, et al. A novel deep learning network with deformable convolution and attention mechanisms for complex scenes ship detection in SAR images[J]. Remote Sensing, 2023, 15(10) :2589-2602.

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献