基于卷积注意力和胶囊网络的 SAR少样本目标识别方法

doi:10.7523/j.ucas.2021.0022

摘要/Abstract

摘要： 合成孔径雷达( SAR)目标识别在军事和民用领域都具有重要的研究价值。但由于SAR数据获取成本高、样本数目少，传统的卷积神经网络提取目标特征的能力不足，准确率低下。提出结合卷积注意力和胶囊网络的分类模型，利用胶囊网络中的多维向量神经元表示目标更多的特征；同时，考虑到少样本情况下目标特征信息缺乏，为提高神经网络的学习效率，对胶囊网络加入注意力机制，通过学习不同特征的重要程度，引导分类网络重点关注对分类结果贡献大的特征，弱化对分类结果贡献小的特征，提高神经网络的学习效率。针对MSTAR数据集和实测车辆数据集的实验结果表明，该算法的准确率高于传统的卷积神经网络和胶囊网络算法。

关键词: 合成孔径雷达, SAR少样本目标识别, 胶囊网络, 卷积神经网络, 卷积注意力机制, 目标检测

Abstract: Synthetic aperture radar (SAR) target recognition has important research value in both military and civil fields. However, due to the high cost of SAR data acquisition and the small number of samples, the traditional convolutional neural network has insufficient ability to extract target features and low accuracy. This paper proposes a classification model combining convolutional attention and capsule network, which uses multi-dimensional vector neurons in the capsule network to represent more features of the target. At the same time, considering the lack of target feature information under the condition of small sample information, in order to improve the learning efficiency of the neural network, attention mechanism is added to the capsule network to guide the classification network to repeat the classification by learning the importance of different features, focusing on the features that contribute more to the classification results, and weakening the features that contribute less to the classification results. The experimental results on MSTAR data set and real vehicle data set show that the accuracy of the proposed algorithm is higher than those of the traditional convolution neural network and capsule network algorithm.

Key words: synthetic aperture radar (SAR), SAR small sample target recognition, capsule network, convolutional neural network, convolutional attention mechanism, object detection

中图分类号:

TN958
TP183

霍鑫怡, 李焱磊, 陈龙永, 张福博, 孙巍. 基于卷积注意力和胶囊网络的 SAR少样本目标识别方法[J]. 中国科学院大学学报, 2022, 39(6): 783-792.

HUO Xinyi, LI Yanlei, CHEN Longyong, ZHANG Fubo, SUN Wei. SAR few-sample target recognition method based on convolutional block attention module and capsule network[J]. Journal of University of Chinese Academy of Sciences, 2022, 39(6): 783-792.

参考文献

[1] 路晋生，张杰敏. 合成孔径雷达的发展历史及趋势[J]. 华北工学院测试技术学报，2000，14(2):80-86.DOI:10.3969/j.issn.1671-7449.2000.02.011.
[2] 贺丰收, 何友, 刘准钆, 等. 卷积神经网络在雷达自动目标识别中的研究进展[J]. 电子与信息学报, 2020, 42(1): 119-131. DOI:10.11999/JEIT180899.
[3] Gu J X, Wang Z H, Kuen J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 354-377. DOI:10.1016/j.patcog.2017.10.013.
[4] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6): 84-90. DOI:10.1145/3065386.
[5] LeCun Y, Boser B, Denker J S, et al. Backpropagation applied to handwritten zip code recognition[J]. Neural Computation, 1989, 1(4): 541-551. DOI:10.1162/neco.1989.1.4.541.
[6] 潘宗序, 安全智, 张冰尘. 基于深度学习的雷达图像目标识别研究进展[J]. 中国科学: 信息科学, 2019, 49(12): 1626-1639.
[7] Ding J, Chen B, Liu H W, et al. Convolutional neural network with data augmentation for SAR target recognition[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(3): 364-368. DOI:10.1109/LGRS.2015.2513754.
[8] Guo J Y, Lei B, Ding C B, et al. Synthetic aperture radar image synthesis by using generative adversarial nets[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(7): 1111-1115. DOI:10.1109/LGRS.2017.2699196.
[9] 王泽隆，徐向辉，张雷. 基于仿真SAR图像深度迁移学习的自动目标识别[J]. 中国科学院大学学报, 2020, 37(4): 516-524. DOI:10.7523/j.issn.2095-6134.2020.04.011.
[10] Huang Z L, Pan Z X, Lei B. What, where, and how to transfer in SAR target recognition based on deep CNNs[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(4): 2324-2336. DOI:10.1109/TGRS.2019.2947634.
[11] 李松，魏中浩，张冰尘, 等. 深度卷积神经网络在迁移学习模式下的SAR目标识别[J]. 中国科学院大学学报, 2018, 35(1): 75-83. DOI:10.7523/j.issn.2095-6134.2018.01.010.
[12] Pan Z X, Bao X J, Zhang Y T, et al. Siamese network based metric learning for SAR target classification[C]//IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium. July 28-August 2, 2019, Yokohama, Japan. IEEE, 2019: 1342-1345. DOI:10.1109/IGARSS.2019.8898210.
[13] Salimans T, Goodfellow I, Zaremba W, et al. Improved techniques for training GANs[C]//NIPS’16: Proceedings of the 30th International Conference on Neural Information Processing Systems(NIPS 16). 2016: 2234-2242.
[14] 邹浩，林赟，洪文. 基于多角度合成SAR图像的目标识别性能分析[J]. 中国科学院大学学报, 2019, 36(2): 226-234. DOI:10.7523/j.issn.2095-6134.2019.02.010.
[15] Chen S Z, Wang H P, Xu F, et al. Target classification using the deep convolutional networks for SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(8): 4806-4817. DOI:10.1109/TGRS.2016.2551720.
[16] Sabour S, Frosst N, Hinton G E. Dynamic routing between capsules[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS17). Long Beach, CA, USA. 3859-3869.
[17] Paoletti M E, Haut J M, Fernandez-Beltran R, et al. Capsule networks for hyperspectral image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(4): 2145-2160. DOI:10.1109/TGRS.2018.2871782.
[18] Guo Y J, Liao J J, Shen G Z. A deep learning model with capsules embedded for high-resolution image classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 14: 214-223. DOI:10.1109/JSTARS.2020.3032672.
[19] 张盼盼, 罗海波, 鞠默然, 等. 一种改进的Capsule及其在SAR图像目标识别中的应用[J]. 红外与激光工程, 2020, 49(5): 203-210. DOI:10.3788/IRLA20201010.
[20] 冯伟业，廖可非，欧阳缮，等. 基于胶囊神经网络的合成孔径雷达图像分类方法[J]. 科学技术与工程，2019，19(28): 203-207.
[21] Xu K, Ba J L, Kiros R, et al. Show, attend and tell: neural image caption generation with visual attention[C]//International Conference on Machine Learning. PMLR, 2015: 2048-2057.
[22] Jaderberg M, Simonyan K, Zisserman A, et al. Spatial transformer networks[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems-Volume 2. 2015: 2017-2025.
[23] Hu J, Shen L, Albanie S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011-2023. DOI:10.1109/TPAMI.2019.2913372.
[24] Woo S, Park J, Lee J Y, et al. CBAM: convolutional block attention module[M]//Computer Vision-ECCV 2018. Cham: Springer International Publishing, 2018: 3-19. DOI:10.1007/978-3-030-01234-2_1.
[25] Ba J, Grosse R, Salakhutdinov R, et al. Learning wake-sleep recurrent attention models[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems-Vol 2. 2015: 2593-2601.
[26] 李红艳, 李春庚, 安居白, 等. 注意力机制改进卷积神经网络的遥感图像目标检测[J]. 中国图象图形学报, 2019, 24(8): 1400-1408.
[27] Wang D, Gao F, Dong J Y, et al. Change detection in synthetic aperture radar images based on convolutional block attention module[C]//2019 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images (MultiTemp). August 5-7, 2019, Shanghai, China. IEEE, 2019: 1-4. DOI:10.1109/Multi-Temp.2019.8866962.
[28] 周龙, 王伟强, 吕科. FACR: 一种快速且准确的车辆识别器[J]. 中国科学院大学学报, 2021, 38(1): 130-136. DOI:10.7523/j.issn.2095-6134.2021.01.016.