基于MFFDet-R的多源舰船图像融合检测方法

doi:10.12305/j.issn.1001-506X.2025.02.06

摘要/Abstract

摘要：

针对对无人机采集到的多源图像的舰船目标融合检测问题, 提出一种基于多模态特征融合旋转检测网络(multi-modal feature fusion detection network based on rotation, MFFDet-R)的多源舰船图像融合检测方法。首先, 为提升检测速度, 采用单阶段无锚框设计降低计算量。随后, 为提升检测精度, 采用旋转任务对齐学习进行标签分配和对齐。然后, 为实现多模态特征的充分融合, 设计多模态特征融合网络。最后, 根据特定场景有针对性地设计检测头和角度预测头, 以提升网络检测性能。通过实验对比验证, 结果表明所提方法可以有效实现对多源舰船的融合检测, 且对不同场景舰船目标的检测性能优于其他方法。

关键词: 多源图像, 融合检测, 任务对齐学习, 特征融合

Abstract:

A multi-source ship image fusion detection method based on multi-modal feature fusion detection network based on rotation (MFFDet-R) is proposed to address the issue of ship target fusion detection for multi-source images obtained by unmanned aerial vehicles. Firstly, a single-stage anchor free frame design is adopted to reduce computational complexity to improve detection speed. Subsequently, rotation task alignment learning is adopted for label allocation and alignment to improve detection accuracy. Then, a multimodal feature fusion network is designed to achieve full fusion of multimodal features. Finally, detection heads and angle prediction heads are designed for specific scenarios to improve network detection performance. Through experimental comparison and verification, the results show that the proposed method can effectively achieve fusion detection of multi-source ships, and its detection performance for ship targets in different scenarios is superior to other methods.

Key words: multi-source image, fusion detection, task alignment learning, feature fusion

中图分类号:

TP391.4

姜杰, 凌青, 闫文君, 刘凯. 基于MFFDet-R的多源舰船图像融合检测方法[J]. 系统工程与电子技术, 2025, 47(2): 390-397.

Jie JIANG, Qing LING, Wenjun YAN, Kai LIU. Multi-source ship image fusion detection method based on MFFDet-R[J]. Systems Engineering and Electronics, 2025, 47(2): 390-397.

图/表 13

图1

图2

图3

图4

表1

表2

图5

表3

表4

表5

图6

图7

图8

参考文献 30

1	王彦情, 马雷, 田原. 光学遥感图像舰船目标检测与识别综述[J]. 自动化学报, 2011, 37 (9): 1029- 1039.
	WANG Y Q , MA L , TIAN Y . Overview of ship target detection and recognition in optical remote sensing images[J]. Journal of Automation, 2011, 37 (9): 1029- 1039.
2	赵其昌, 吴一全, 苑玉彬. 光学遥感图像舰船目标检测与识别方法研究进展[J]. 航空学报, 2023, 34 (1): 242- 251.
	ZHAO Q C , WU Y Q , YUAN Y B . Research progress on ship target detection and recognition methods in optical remote sensing images[J]. Journal of Aeronautics, 2023, 34 (1): 242- 251.
3	何友, 熊伟, 刘俊, 等. 海上信息感知与融合研究进展及展望[J]. 火力与指挥控制, 2018, 43 (6): 1- 10. doi: 10.3969/j.issn.1002-0640.2018.06.001
	HE Y , XIONG W , LIU J , et al. Research progress and prospects on maritime information perception and fusion[J]. Firepower and Command and Control, 2018, 43 (6): 1- 10. doi: 10.3969/j.issn.1002-0640.2018.06.001
4	甘春生. 星载遥感图像舰船检测方法研究[D]. 辽宁: 沈阳航空航天大学, 2016.
	GAN C S. Research on ship detection methods in spaceborne remote sensing images[D]. Liaoning: Shenyang University of Aeronautics and Astronautics, 2016.
5	LEI S , ZOU Z X , LIU D G , et al. Sea-land segmentation for infrared remote sensing images based on superpixels and multi-scale features[J]. Infrared Physics & Technology, 2018, 91, 12- 17.
6	ZHI B X , ZHOU F . Analysis of new top-hat transformation and the application for infrared dim small target detection[J]. Pattern Recognition, 2010, 43 (6): 2145- 2156. doi: 10.1016/j.patcog.2009.12.023
7	LIU R , LU Y , GONG C , et al. Infrared point target detection with improved template matching[J]. Infrared Physics & Technology, 2012, 55 (4): 380- 387.
8	李海军, 孔繁程, 林云. 基于改进YOLOv5s的红外舰船检测算法[J]. 系统工程与电子技术, 2023, 45 (8): 2415- 2422.
	LI H J , KONG F C , LIN Y . Infrared ship detection algorithm based on improved YOLOv5s[J]. Systems Engineering and Electronics, 2023, 45 (8): 2415- 2422.
9	潘为年. 基于深度学习的红外成像舰船目标检测方法研究[D]. 成都: 电子科技大学, 2021.
	PAN W N. Research on infrared imaging ship target detection method based on deep learning[D]. Chengdu: University of Electronic Science and Technology of China, 2021.
10	ZHANG J X . Multi-source remote sensing data fusion: status and trends[J]. International Journal of Image and Data Fusion, 2010, 1 (1): 5- 24. doi: 10.1080/19479830903561035
11	LI M J, DONG Y B, WANG X L. Pixel level image fusion based the wavelet transform[C]//Proc. of the 6th International Congress on Image and Signal Processing, 2013, 2: 995-999.
12	HAN X , LYU Y , SONG T X . An adaptive two-scale image fusion of visible and infrared images[J]. IEEE Access, 2019, 7, 56341- 56352. doi: 10.1109/ACCESS.2019.2913289
13	YOU T T, TANG Y. Visual saliency detection based on adaptive fusion of color and texture features[C]//Proc. of the 3rd IEEE International Conference on Computer and Communications, 2017: 2034-2039.
14	杨曦, 张鑫, 郭浩远, 等. 基于不变特征的多源遥感图像舰船目标检测算法[J]. 电子学报, 2022, 50 (4): 887.
	YANG X , ZHANG X , GUO H Y , et al. Ship target detection algorithm based on invariant features in multi-source remote sensing images[J]. Journal of Electronics, 2022, 50 (4): 887.
15	WANG A, JIANG J N, ZHANG H Y. Multi-sensor image decision level fusion detection algorithm based on D-S evidence theory[C]//Proc. of the 4th International Conference on Instrumentation and Measurement, Computer, Communication and Control, 2014, 620-623.
16	PAUL P P , GAVRILOVA M L , ALHAJJ R . Decision fusion for multimodal biometrics using social network analysis[J]. IEEE Trans. on Systems, Man, and Cybernetics: Systems, 2014, 44 (11): 1522- 1533. doi: 10.1109/TSMC.2014.2331920
17	关欣, 国佳恩, 衣晓. 基于低秩双线性池化注意力网络的舰船目标识别[J]. 系统工程与电子技术, 2023, 45 (5): 1305- 1314.
	GUAN X , GUO J E , YI X . Ship target recognition based on low rank bilinear pooling attention network[J]. Systems Engineering and Electronics, 2023, 45 (5): 1305- 1314.
18	DELIANG X , YIHAO X U , JIANDA C , et al. An algorithm based on a feature interaction-based keypoint detector and sim-CSPNet for SAR image registration[J]. Journal of Radars, 2022, 11 (6): 1081- 1097.
19	ZHANG Y C , ZHANG W B , YU J Y , et al. Complete and accurate holly fruits counting using YOLOX object detection[J]. Computers and Electronics in Agriculture, 2022, 198, 107062. doi: 10.1016/j.compag.2022.107062
20	DEVER W G . The chronology of Syria-Palestine in the second millennium BCE: a review of current issues[J]. Bulletin of the American Schools of Oriental Research, 1992, 288 (1): 1- 25.
21	SONG G L, LIU Y, WANG X G. Revisiting the sibling head in object detector[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 11563-11572.
22	FU A M , ZHANG X L , XIONG N X , et al. VFL: a verifiable federated learning with privacy-preserving for big data in industrial IOT[J]. IEEE Trans. on Industrial Informatics, 2020, 18 (5): 3316- 3326.
23	FENG C J, ZHONG Y J, GAO Y, et al. Tood: task-aligned one-stage object detection[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2021: 3490-3499.
24	LI X , WANG W H , WU L J , et al. Generalized focal loss: learning qualified and distributed bounding boxes for dense object detection[J]. Advances in Neural Information Processing Systems, 2020, 33, 21002- 21012.
25	LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proc. of the IEEE conference on computer vision and pattern recognition, 2017: 2117-2125.
26	LIU Z, LI Y, YAO L, et al. Task aligned generative meta-learning for zero-shot learning[C]//Proc. of the AAAI Confe-rence on Artificial Intelligence, 2021, 35(10): 8723-8731.
27	LLERENA J E. Probabilistic intersection-over-union for training and evaluation of oriented object detectors[J]. 2022, 15(6): 156-178.
28	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
29	ZHANG H Y, WANG Y, DAYOUB F, et al. Varifocalnet: an IoU-aware dense object detector[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 8514-8523.
30	LI X , WANG W H , WU L J , et al. Generalized focal loss: learning qualified and distributed bounding boxes for dense object detection[J]. Advances in Neural Information Processing Systems, 2020, 33, 21002- 21012.

模型	mAP/%	参数量(MFLOPs)	计算量(GFLOPs)
基准	75.61	50.65	269.09
+旋转任务对齐学习	77.24(+1.63)	50.65	269.09
+解耦角度预测头	77.78(+0.54)	52.20	272.72
+角度预测与DFL	78.01(+0.23)	53.29	281.65
+可学习门控单元	78.14(+0.13)	53.29	281.65

算法	mAP1/%	mAP2/%	FPS
单发对齐网络	79.4	58.7	70.3
基于特征精炼的单阶段检测网络	76.5	57.4	65.1
定向旋转检测网络	80.8	59.3	66.7
基于点集表示的旋转目标检测网络	77.7	58.3	72.3
RoI转换器网络	80.9	59.8	62.2
无锚框任意朝向船舶目标检测网络	76.1	54.7	59.8
本文网络	82.7	62.3	78.0

函数	mAP
高斯损失函数	78.35
旋转目标检测损失函数	75.61
相对熵	77.32

模态	mAP
可见光	82.5
红外	75.2
红外+红外	75.4
可见光+可见光	82.7
可见光+红外	90.2

算法	mAP1	mAP2
YOLOX	93.4	80.1
YOLOV7	93.5	81.2
可扩展高效目标检测网络	92.1	80.6
单发多盒检测网络	90.5	79.8
区域卷积神经网络	91.4	79.9
视网膜网络	88.9	78.7
全卷积单阶段目标检测	89.2	78.3
本文网络	93.8	83.4