基于混合型复数域卷积神经网络的三维转动舰船目标识别

doi:10.12263/DZXB.20211255

电子学报 ›› 2022, Vol. 50 ›› Issue (5): 1042-1049.DOI: 10.12263/DZXB.20211255

所属专题：长摘要论文

基于混合型复数域卷积神经网络的三维转动舰船目标识别

张云, 化青龙, 姜义成, 徐丹

哈尔滨工业大学电子与信息工程学院，黑龙江哈尔滨 150001

收稿日期:2021-09-13 修回日期:2022-02-20 出版日期:2022-05-25
- 通讯作者:
- 姜义成
- 作者简介:
- 张云女，1975年11月出生于黑龙江省虎林市.现为哈尔滨工业大学教授、博士生导师.主要研究方向为雷达信号处理、SAR成像、机器学习和遥感模式分析.E-mail: zhangyunhit@hit.edu.cn
  化青龙男，1995年2月出生于安徽省阜阳市.现为哈尔滨工业大学博士.主要研究方向为雷达图像处理和深度学习网络框架.E-mail: huaqinglong_hit@163.com
  姜义成男，1964年11月出生于黑龙江省哈尔滨市.现为哈尔滨工业大学教授、博士生导师.主要研究方向为雷达信号处理.E-mail:jiangyc@hit. edu. cn
  徐丹女，1996年4月出生于黑龙江省七台河市.主要研究方向为数据分析与智能图像处理.E-mail: xudanhit@hit.edu.cn
- 基金资助:
- 国家自然科学基金 (61201304)

Recognition of 3D Rotating Ship Based on Mix-CV-CNN

ZHANG Yun, HUA Qing-long, JIANG Yi-cheng, XU Dan

School of Electronics and Information Engineering, Harbin Institute of Technology, Harbin, Heilongjiang 150001, China

Received:2021-09-13 Revised:2022-02-20 Online:2022-05-25 Published:2022-06-18
- Corresponding author:
- JIANG Yi-cheng
- Supported by:
- National Natural Science Foundation of China (61201304)

摘要/Abstract

摘要：

在较高海情下，由于舰船目标处于随机摆动的非平稳运动状态，常规合成孔径雷达（Synthetic Aperture Radar， SAR）成像处理会使得目标散焦、方位模糊，从而导致三维转动舰船目标识别准确率低.本文提出一种混合型复数域卷积神经网络（Mix-type Complex-Valued Convolutional Neural Network，Mix-CV-CNN），并推导Mix-CV-CNN前向传播与反向传播算法.三维转动舰船目标经过SAR成像处理后存在剩余相位信息，Mix-CV-CNN能充分利用SAR复数域图像的幅度和相位信息，在不进行目标重聚焦的情况下，较好完成SAR复杂运动舰船目标的识别.实验表明，Mix-CV-CNN相较于具有相同自由度的实数域卷积神经网络（Real-Valued Convolutional Neural Network，RV-CNN）识别性能有所提高，实测数据识别平均准确率提高3.85%.

关键词: 合成孔径雷达, 复数域卷积神经网络, 三维转动, 目标散焦, 舰船目标识别, 混合型复数域卷积神经网络

Abstract:

Because the ship targets are in a non-stationary motion state of random swing, conventional synthetic aperture radar(SAR) imaging processing will make the targets defocused and azimuth blurred, resulting in the recognition accuracy of three-dimensional rotating ship. This paper proposes a mixed-type complex-valued convolutional neural network(Mix-CV-CNN) and derives the Mix-CV-CNN forward propagation and backpropagation algorithms. The three-dimensional rotating target has residual phase information after SAR imaging processing. The Mix-CV-CNN could make full use of the amplitude and phase information of the complex SAR image and could better complete the recognition of SAR three-dimensional rotating targets without target refocusing. The experimental results show that Mix-CV-CNN has improved recognition performance compared with the real-valued convolutional neural network(RV-CNN) with the same degree of freedom. The average accuracy is increased by 3.85%.

Key words: synthetic aperture radar(SAR), complex-valued convolutional neural network(CV-CNN), three-dimensional rotation, target defocus, ship target classification, mix-type complex-valued convolutional neural network(Mix-CV-CNN)

中图分类号:

TP751

张云, 化青龙, 姜义成, 徐丹. 基于混合型复数域卷积神经网络的三维转动舰船目标识别[J]. 电子学报, 2022, 50(5): 1042-1049.

ZHANG Yun, HUA Qing-long, JIANG Yi-cheng, XU Dan. Recognition of 3D Rotating Ship Based on Mix-CV-CNN[J]. Acta Electronica Sinica, 2022, 50(5): 1042-1049.

图/表 12

参考文献 21

1	ZHAOY, ZHANGM, ZHAOY, GENGX. A bistatic SAR image intensity model for the composite ship-ocean scene[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(8): 4250-4258.
2	LIUP, JINY. A study of ship rotation effects on SAR image[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(6): 3132-3144.
3	NEWEYM, BENITZG R, BARRETTD J, et al. Detection and imaging of moving targets with LiMIT SAR data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(6): 3499-3510.
4	KRIZHEVSKYA, SUTSKEVERI, HINTONG E. Imagenet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25: 1097-1105.
5	张晓玲, 张天文, 师君, 等. 基于深度分离卷积神经网络的高速高精度SAR舰船检测[J]. 雷达学报, 2019, 8(6): 841-851.
	ZHANGXiao-ling, ZHANGTian-wen, SHIJun, et al. High-speed and High-accurate SAR ship detection based on a depthwise separable convolution neural network[J]. Journal of Radars, 2019, 8(6): 841-851. (in Chinese)
6	杨龙, 苏娟, 李响. 基于深度卷积神经网络的SAR舰船目标检测[J]. 系统工程与电子技术, 2019, 41(09): 1990-1997.
	YANGLong, SUJuan, LIXiang. Ship detection in SAR images based on deep convolutional neural network[J]. Systems Engineering and Electronics, 2019, 41(09): 1990-1997. (in Chinese)
7	杜兰, 王兆成, 王燕, 等. 复杂场景下单通道SAR目标检测及鉴别研究进展综述[J]. 雷达学报, 2020, 9(01): 34-54.
	DULan, WANGZhao-cheng, WANGYan, et al. Survey of research progress on target detection and discrimination of single-channel SAR images for complex scenes[J]. Journal of Radars, 2020, 9(01): 34-54. (in Chinese)
8	ZHANGZ, WANGH, XUF, et al. Complex- valued convolutional neural network and its application in polarimetric SAR image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(12): 7177-7188.
9	YUL, HUY, XIEX, et al. Complex-valued full convolutional neural network for SAR target classification[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(10): 1752-1756.
10	TANX, LIM, ZHANGP, et al. Complex-valued 3-D convolutional neural network for PolSAR image classification[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(6): 1022-1026.
11	SUNAGAY, NATSUAKIR, HIROSEA. Land form classification and similar land-shape discovery by using complex-valued convolutional neural networks[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(10): 7907-7917.
12	GAOJ, DENGB, QINY, et al. Enhanced radar imaging using a complex-valued convolutional neural network[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(1): 35-39.
13	蔡彦宁, 种劲松. 舰船目标三维转动SAR成像仿真研究[J]. 计算机仿真, 2011, 28(03): 287-291, 364.
	CAIYan-ning, ZHONGJin-song. SAR imaging simulation for three-dimensional rotations of ship targets[J]. Computer Simulation, 2011, 28(03): 287-291, 364. (in Chinese)
14	KUROEY, YOSHIDM, MORIT. On activation functions for complex-valued neural networks-existence of energy functions[C]//Artificial Neural Networks and Neural Information Processing. Heidelberg, Berlin: Springer, 2003: 985-992.
15	HIROSEA. Complex-valued neural networks: theories and applications[M]. Singapore: World Scientific, 2003.
16	ILIEVA, KYURKCHIEVN, MARKOVS. On the approximation of the step function by some sigmoid functions[J]. Mathematics and Computers in Simulation, 2017, 133:223-234.
17	NAIRV, HINTONG E. Rectified linear units improve restricted Boltzmann machines[C]//27th International Conference on Machine Learning. Haifa, Israel: ICML, 2010: 1-8.
18	项鹏, 郭炜炜, 张增辉,等. 面向SAR解译的OpenSAR数据开放共享平台[J]. 信息技术, 2016, (09): 1-4, 9.
	XIANGPeng, GUOWei-wei, LIZeng-hui, et al. An open platform of SAR database for SAR image interpretation[J]. Information Technology, 2016, (09): 1-4, 9. (in Chinese)
19	ZHANGY, HUAQ, XUD, et al. A complex-valued convolutional neural network with different activation functions in polarimetric SAR image classification[C]// 2019 International Radar Conference. Toulon, France: IEEE, 2019: 1-4.
20	SIMONYANK, ZISSERMANA. Very deep convolutional networks for large-scale image recognition[EB/OL]. [2014]. .
21	HEK, ZHANGX, RENS, et al. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, America: IEEE, 2016: 770-778.

类别	Mix-CV-CNN	RI-CV-CNN	RV-CNN	VGG16	ResNet18
阿利伯克级	99.36	96.21	95.31	99.92	99.14
黑天鹅级	97.76	92.49	80.76	95.52	88.36
不来梅级	96.43	91.33	80.35	88.61	78.12
日向级	96.54	95.29	93.90	94.51	96.74
罗津级	97.32	94.24	75.67	85.49	96.42
平均准确率	97.46	93.91	85.45	92.81	91.76

类别	Mix-CV-CNN	RI-CV-CNN	RV-CNN	VGG16	ResNet18
阿利伯克级	99.36	96.21	95.31	99.92	99.14
黑天鹅级	97.76	92.49	80.76	95.52	88.36
不来梅级	96.43	91.33	80.35	88.61	78.12
日向级	96.54	95.29	93.90	94.51	96.74
罗津级	97.32	94.24	75.67	85.49	96.42
平均准确率	97.46	93.91	85.45	92.81	91.76

	1	2	3	4	5
1	99.36	0	0	0.64	0
2	0	97.76	0.22	0	2.01
3	0	0	96.43	0	3.57
4	0.41	0	3.05	96.54	0
5	1.56	0.67	0.45	0	97.32

	1	2	3	4	5
1	99.36	0	0	0.64	0
2	0	97.76	0.22	0	2.01
3	0	0	96.43	0	3.57
4	0.41	0	3.05	96.54	0
5	1.56	0.67	0.45	0	97.32

算法	总耗时(s)	每批次耗时(s)	参数量(M)
Mix-CV-CNN	62	0.06	9.65
RI-CV-CNN	59	0.06	9.65
RV-CNN	21	0.02	4.83
VGG16	141	0.14	24.15
ResNet18	139	0.14	11.51

基于混合型复数域卷积神经网络的三维转动舰船目标识别

Recognition of 3D Rotating Ship Based on Mix-CV-CNN

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 21

相关文章 15

编辑推荐

Metrics

类别	Mix-CV-CNN	RI-CV-CNN	RV-CNN	VGG16	ResNet18
货船	97.14	93.56	94.29	93.54	91.93
油船	98.08	94.17	84.62	89.36	97.87
其他船	93.10	95.44	96.55	86.27	94.11
平均准确率	96.09	94.39	92.24	89.72	94.63

	货船	油船	其他船
货船	97.14	0	2.86
油船	1.92	98.08	0
其他船	0	6.90	93.10

[1]	张继龙, 李业振, 刘勇, 王栋. 基于透镜成像原理的微波阵列快速成像技术[J]. 电子学报, 2022, 50(5): 1107-1116.
[2]	李玉, 王姝运, 赵泉华. 基于斑点统计特性保持的SAR影像迭代滤波[J]. 电子学报, 2021, 49(9): 1809-1817.
[3]	王明春, 张嘉峰, 杨子渊, 刘涛. Beta分布下基于白化滤波的极化SAR图像海面舰船目标CFAR检测方法[J]. 电子学报, 2019, 47(9): 1883-1890.
[4]	张嘉峰, 杨子渊, 张鹏, 刘涛. Fisher纹理分布下基于匹配滤波的极化SAR图像CFAR检测方法[J]. 电子学报, 2019, 47(12): 2533-2543.
[5]	李亮, 洪峻, 陈琦, 王爱春, 王宇, 明峰, 朱勇涛. 基于极化有源定标器的高分三号SAR在轨测试分析[J]. 电子学报, 2018, 46(9): 2157-2164.
[6]	李东, 赵婷, 宋伟, 刘庆华, 曾浩, 张成祥. 一种低信噪比下稳健的ISAR平动补偿方法[J]. 电子学报, 2018, 46(9): 2049-2056.
[7]	王彬, 王国宇. 基于广义Gamma分布的高分辨率SAR图像海岸线检测[J]. 电子学报, 2018, 46(4): 827-833.
[8]	陈文峰, 李少东, 杨军, 马晓岩. 低信噪比下二维联合快速超分辨B-ISAR成像方法[J]. 电子学报, 2018, 46(4): 840-848.
[9]	张嘉峰, 朱博, 张鹏, 王明春, 刘涛. Wishart分布情形下极化SAR图像目标CFAR检测解析方法[J]. 电子学报, 2018, 46(2): 433-439.
[10]	张嘉峰, 张鹏, 王明春, 刘涛. Fisher分布下基于白化滤波的极化SAR图像CFAR检测方法[J]. 电子学报, 2018, 46(12): 2854-2861.
[11]	张云鹏, 毕大平, 房明星, 沈爱国. 对SAR双通道对消的方位向间歇采样散射波干扰[J]. 电子学报, 2018, 46(12): 2832-2839.
[12]	王成浩, 廖桂生, 许京伟. FDA-SAR高分辨宽测绘带成像距离解模糊方法[J]. 电子学报, 2017, 45(9): 2085-2091.
[13]	李东, 占木杨, 粟嘉, 方志平, 曹海林. 一种基于相干积累CPF和NUFFT的机动目标ISAR成像新方法[J]. 电子学报, 2017, 45(9): 2225-2232.
[14]	卜红霞, 白霞, 赵娟, 齐耀辉, 闫若颖. 基于压缩感知的矩阵型联合SAR成像与自聚焦算法[J]. 电子学报, 2017, 45(4): 874-881.
[15]	李少东, 陈文峰, 杨军, 马晓岩, 吕明久. 稀疏孔径下的运动补偿及快速超分辨成像方法[J]. 电子学报, 2017, 45(2): 291-299.