基于CycleGAN的月表图像数据增强方法

doi:10.12305/j.issn.1001-506X.2023.10.06

系统工程与电子技术 ›› 2023, Vol. 45 ›› Issue (10): 3041-3048.doi: 10.12305/j.issn.1001-506X.2023.10.06

• 电子技术 • 上一篇

基于CycleGAN的月表图像数据增强方法

宋婷¹^,², 兀泽朝³^,⁴, 高艾⁵^,*, 袁建平¹

1. 西北工业大学航天学院, 陕西西安 710072
2. 上海航天控制技术研究所, 上海 201109
3. 北京机电工程研究所, 北京 100074
4. 复杂系统控制与智能协同技术重点实验室, 北京 100074
5. 北京理工大学宇航学院, 北京 100081

收稿日期:2022-07-13 出版日期:2023-09-25 发布日期:2023-10-11
通讯作者: 高艾
作者简介:宋婷(1984—), 女, 高级工程师, 博士研究生, 主要研究方向为航天器规划与控制
兀泽朝(1997—), 男, 助理工程师, 硕士, 主要研究方向为飞行器自主导航、制导与控制
高艾(1984—), 女, 教授, 博士, 主要研究方向为航天器自主导航与制导控制
袁建平(1957—), 男, 教授, 博士, 主要研究方向为航天器动力学
基金资助:
国家自然科学基金(11872110)

CycleGAN-based data enhancement method for lunar surface images

Ting SONG¹^,², Zezhao WU³^,⁴, Ai GAO⁵^,*, Jianping YUAN¹

1. School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China
2. Shanghai Aerospace Control Technology Institute, Shanghai 201109, China
3. Institute of Mechanical and Electrical Engineering, Beijing 100074, China
4. Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory, Beijing 100074, China
5. School of Aerospace Engineering, Beijing Institute of Technology, Beijing 100081, China

Received:2022-07-13 Online:2023-09-25 Published:2023-10-11
Contact: Ai GAO

摘要/Abstract

摘要：

针对月球表面先验图像信息获取困难的问题, 提出一种基于对抗神经网络的月表先验图像数据增强方法。在获取少量月表图像及障碍背景分割图的基础上, 构建基于对抗神经网络的月表图像数据增强架构, 利用新的障碍背景分割图匹配生成月表图像, 扩充月表先验图像数据, 可用于月球探测中障碍检测算法设计验证。仿真结果证明了所提方法生成的月表图像接近真实拍摄图像, 且通过数据增强图像数据, 使障碍检测结果获得明显提升, 证明了方法的有效性。

关键词: 月球探测, 数据增强, 深度学习, 对抗神经网络

Abstract:

In this paper, a method of lunar surface priori image data enhancement based on adversarial neural network is proposed to address the problem of difficulty in acquiring a priori image information on the lunar surface. Based on the acquisition of a small amount of lunar surface images and obstacle background segmentation maps, the lunar surface image data enhancement architecture based on the adversarial neural network is constructed, and the new obstacle background segmentation maps are used to match the lunar surface images and expand the lunar surface priori image data, which can be used for the design and verification of obstacle detection algorithms in lunar exploration. Simulation results prove that the lunar surface images generated by the proposed method are close to the real captured images, and the image data is enhanced by the data to obtain obvious improvement of the obstacle detection results, which proves the effectiveness of the proposed method.

Key words: lunar exploration, data enhancement, deep learning, adversarial neural networks

中图分类号:

V448

宋婷, 兀泽朝, 高艾, 袁建平. 基于CycleGAN的月表图像数据增强方法[J]. 系统工程与电子技术, 2023, 45(10): 3041-3048.

Ting SONG, Zezhao WU, Ai GAO, Jianping YUAN. CycleGAN-based data enhancement method for lunar surface images[J]. Systems Engineering and Electronics, 2023, 45(10): 3041-3048.

图/表 15

图1

图2

图3

图4

图5

图6

图7

表1

图8

图9

表2

图10

图11

图12

表3

参考文献 30

1	崔平远, 高艾, 朱圣英. 深空探测器自主导航与制导[M]. 北京: 中国宇航出版社, 2016: 4- 10.
	CUI P Y , GAO A , ZHU S Y . Autonomous navigation and gui-dance of deep space probe[M]. Beijing: China Astronautic Publishing House, 2016: 4- 10.
2	宁晓琳, 蔡洪炜, 吴伟仁, 等. 月球车的惯性/天文组合导航新方法[J]. 系统工程与电子技术, 2011, 33 (8): 1837- 1844.
	NING X L , CAI H W , WU W R , et al. INS/CNS integrated navigation method for lunar rover[J]. Journal of Systems Engineering and Electronics, 2011, 33 (8): 1837- 1844.
3	吴伟仁, 于登云. 深空探测发展与未来关键技术[J]. 深空探测学报(中英文), 2014, 1 (1): 5- 17.
	WU W R , YU D Y . Development of deep space exploration and its future key technologies[J]. Journal of Deep Space Exploration, 2014, 1 (1): 5- 17.
4	于正湜, 崔平远. 行星着陆自主导航与制导控制研究现状与趋势[J]. 深空探测学报, 2016, 3 (4): 345- 355.
	YU Z S , CUI P Y . Research status and developing trend of the autonomous navigation, guidance, and control for planetary landing[J]. Journal of Deep Space Exploration, 2016, 3 (4): 345- 355.
5	WU W R , LIU W W , QIAO D , et al. Investigation on the development of deep space exploration[J]. Science China Technological Sciences, 2012, 55 (4): 1086- 1091. doi: 10.1007/s11431-012-4759-z
6	RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[EB/OL]. [2022-11-17]. https://arxiv.org/abs/1511.06434.
7	NG T T, CHANG S F, SUN Q. A data set of authentic and spliced image blocks[EB/OL]. [2022-11-17]. https://arxiv.org/abs/2107.07699v1.
8	JOLY A , GOEAU H , BONNET P , et al. Interactive plant identification based on social image data[J]. Ecological Informa- tics, 2014, 23, 22- 34. doi: 10.1016/j.ecoinf.2013.07.006
9	HAO P, LI C, RAHAMAN M M, et al. A comparison of deep learning classification methods on small-scale image data set: from convolutional neural networks to visual transformers[EB/OL]. [2022-11-17]. https://arxiv.org/abs/2107.07699v1.
10	陈坤, 王璐, 储珺. 月球表面图像的SIFT特征提取与匹配[J]. 计算机与现代化, 2011, (7): 20-23, 26.
	CHEN K , WANG L , CHU J . SIFT feature extraction and matching of lunar surface image[J]. Computer and Modernization, 2011, (7): 20-23, 26.
11	欧阳自远. 月球探测的进展与中国的月球探测[J]. 地质科技情报, 2004, 23 (4): 1- 5.
	OUYANG Z Y . International lunar exploration progress and chinese lunar exploration[J]. Bulletin of Geologic Science and Technology, 2004, 23 (4): 1- 5.
12	秦同, 朱圣英, 崔平远, 等. 行星着陆动力下降段相对视觉导航方法[J]. 宇航学报, 2019, 40 (2): 164- 173.
	QIN T , ZHU S Y , CUI P Y , et al. Relative optical navigation in powered descent phase of planetary landings[J]. Journal of Astronautics, 2019, 40 (2): 164- 173.
13	HE J , WANG C , JIANG D , et al. Cyclegan with an improved loss function for cell detection using partly labeled images[J]. IEEE Journal of Biomedical and Health Informatics, 2020, 24 (9): 2473- 2480.
14	SPEYERER E J, ROBINSON M S, DENEVI B W. Lunar reconnaissance orbiter camera global morphological map of the moon[C]//Proc. of the 42nd Annual Lunar and Planetary Science Conference, 2011, (1608): 2387.
15	GOODFELLOW I J , POUGET-ABADIE J , MIRZA M , et al. Generative adversarial nets[M]. Cambridge: MIT Press, 2014.
16	WANG K F , GOU C , DUAN Y J , et al. Generative adversarial networks: introduction and outlook[J]. IEEE/CAA Journal of Automatica Sinica, 2017, 4 (4): 588- 598.
17	MIRZA M , OSINDERO S . Conditional generative adversarial nets[J]. Computer Science, 2014, 2672- 2680.
18	Z HU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]// Proc. of the IEEE International Conference on Computer Vision, 2017: 2223-2232.
19	CHU C, ZHMOGINOV A, SANDLER M. Cyclegan, a master of steganography[EB/OL]. [2022-07-13]. https://arxiv.org/abs/1712.02950.
20	MAHMOOD F , BORDERS D , CHEN R J , et al. Deep adversarial training for multi-organ nuclei segmentation in histopathology images[J]. IEEE Trans. on Medical Imaging, 2019, 39 (11): 3257- 3267.
21	ALMAHAIRI A, RAJESHWAR S, SORDONI A, et al. Augmented cyclegan: learning many-to-many mappings from unpaired data[C]//Proc. of the International Conference on Machine Learning, 2018: 195-204.
22	KANEKO T, KAMEOKA H, TANAKA K, et al. Cyclegan-vc2: Improved cyclegan-based non-parallel voice conversion[C]// Proc. of the ICASSP IEEE International Conference on Acoustics, Speech and Signal Processing, 2019: 6820-6824.
23	ISOLA P, ZHU J Y, ZHOU T, et al. Image-to-image translation with conditional adversarial networks[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 1125-1134.
24	LIU X, HSIEH C J. Rob-gan: generator, discriminator, and adversarial attacker[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 11234-11243.
25	CHANG H, LU J, YU F, et al. PairedCycleGAN: asymmetric style transfer for applying and removing makeup[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018.
26	YASUNO T, FUJII J, FUKAMI S. One-class steel detector using patch GAN discriminator for visualising anomalous feature map[EB/OL]. [2022-11-17]. https://arxiv.org/abs/2107.00143.
27	DEMIR U, UNAL G. Patch-based image inpainting with ge-nerative adversarial networks[EB/OL]. [2022-11-17]. https://arxiv.org/abs/1803.07422v1.
28	CHOI Y, GHOI M, KIM M, et al. StarGAN: unified generative adversarial networks for multi-domain image-to-image translation[C]//Proc. of the International Conference on Computer Vision and Recognition, 2018: 8789-8797.
29	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proc. of the European Conference on Computer Vision, 2018: 3-19.
30	WANG J S , GAI S , HUANG X , et al. From coarse to fine: a two stage conditional generative adversarial network for single image rain removal[J]. Digital Signal Processing, 2021, 111 (11): 102985.

模块	卷积块	输入特征(b, c, h, w)	卷积层参数(n, k, s, p)	输出特征(b, c, h, w)
输入	-	(1, 3, 256, 256)	-	-
卷积部分Conv	CB1	(1, 3, 256, 256)	(64, 7, 1, 3)	(1, 64, 128, 128)
	CB2	(1, 64, 128, 128)	(128, 3, 2, 1)	(1, 128, 64, 64)
	CB3	(1, 128, 64, 64)	(256, 3, 2, 1)	(1, 256, 32, 32)
残差部分Residual	RB*9	(1, 256, 32, 32)	(256, 3, 1, 1)×2×9	(1, 256, 32, 32)
反卷积部分Deconv	DB3	(1, 256, 32, 32)	(128, 3, 2, 1)	(1, 128, 64, 64)
	DB4	(1, 128, 64, 64)	(64, 3, 2, 1)	(1, 64, 128, 128)
	DB5	(1, 64, 128, 128)	(3, 7, 1, 3)	(1, 3, 256, 256)

卷积层	输入特征(b, c, h, w)	卷积层参数(n, k, s, p)	输出特征(b, c, h, w)
网络输入	(1, 3, 256, 256)	-	-
卷积1	(1, 3, 256, 256)	(64, 4, 2, 1)	(1, 64, 128, 128)
卷积2	(1, 64, 128, 128)	(128, 4, 2, 1)	(1, 128, 64, 64)
卷积3	(1, 128, 64, 64)	(256, 4, 2, 1)	(1, 256, 32, 32)
卷积4	(1, 256, 32, 32)	(512, 4, 1, 1)	(1, 512, 31, 31)
卷积5	(1, 512, 31, 31)	(1, 4, 1, 1)	(1, 1, 30, 30)

评价指标	轨道拍摄器拍摄图像	轨道拍摄器拍摄图像+CycleGAN
准确率	96.9	97.3
查准率	91.2	91.8
查全率	75.6	78.8
F1分数	82.6	84.8

[1]	汪萌, 诸兵. 不确定性建模在2D和3D目标检测中的应用[J]. 系统工程与电子技术, 2023, 45(8): 2370-2376.
[2]	邵凯, 杜自群, 王光宇. 基于模型剪枝动态调整压缩率的CSI反馈方法[J]. 系统工程与电子技术, 2023, 45(8): 2615-2622.
[3]	崔天舒, 王栋, 黄振. 面向空间认知通信的轻量化网络自动调制分类方法[J]. 系统工程与电子技术, 2023, 45(7): 2220-2226.
[4]	姜雨, 袁琪, 胡志韬, 吴薇薇, 顾欣. 基于气象因素的机场进离港延误预测[J]. 系统工程与电子技术, 2023, 45(6): 1722-1731.
[5]	陈洋, 廖灿辉, 张锟, 刘建, 王鹏举. 基于自监督对比学习的信号调制识别算法[J]. 系统工程与电子技术, 2023, 45(4): 1200-1206.
[6]	张晔, 侯毅, 欧阳克威, 周石琳. 单变量序列数据分类方法综述[J]. 系统工程与电子技术, 2023, 45(2): 313-335.
[7]	邵正途, 许登荣, 徐文利, 王晗中. 基于LSTM和残差网络的雷达有源干扰识别[J]. 系统工程与电子技术, 2023, 45(2): 416-423.
[8]	李瑞泽, 张双辉, 刘永祥. 基于卷积ADMM网络的高效结构化稀疏ISAR成像方法[J]. 系统工程与电子技术, 2023, 45(1): 56-70.
[9]	韩啸, 陈世文, 陈蒙, 杨锦程. 基于互易点学习的LPI信号开集识别[J]. 系统工程与电子技术, 2022, 44(9): 2752-2759.
[10]	王彩云, 吴钇达, 王佳宁, 马璐, 赵焕玥. 基于改进的CNN和数据增强的SAR目标识别[J]. 系统工程与电子技术, 2022, 44(8): 2483-2487.
[11]	张立民, 谭凯文, 闫文君, 张聿远. 基于多级跳线残差网络的雷达辐射源识别[J]. 系统工程与电子技术, 2022, 44(7): 2148-2156.
[12]	金国栋, 薛远亮, 谭力宁, 许剑锟. 基于孪生神经网络的目标跟踪算法进展研究[J]. 系统工程与电子技术, 2022, 44(6): 1805-1822.
[13]	赵晓枫, 徐叶斌, 吴飞, 牛家辉, 蔡伟, 张志利. 基于全局感知机制的地面红外目标检测方法[J]. 系统工程与电子技术, 2022, 44(5): 1461-1467.
[14]	邹虹, 白陈阳, 何鹏, 崔亚平, 王汝言, 吴大鹏. 基于分布式深度学习的边缘服务放置策略[J]. 系统工程与电子技术, 2022, 44(5): 1728-1737.
[15]	陈冬, 句彦伟. 基于语义分割实现的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2022, 44(4): 1195-1201.

基于CycleGAN的月表图像数据增强方法

CycleGAN-based data enhancement method for lunar surface images

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 30

相关文章 15

编辑推荐

Metrics

本文评价