一种星载平台轻量化快速影像匹配方法

doi:10.12082/dqxxkx.2022.210529

摘要/Abstract

摘要：

针对现有传统影像匹配在星载平台有限的存储及算力条件下难以快速计算的问题,本文提出一种适用于星载平台的、基于哈希学习的轻量化快速影像匹配方法。该方法以同源卫星影像的特征描述符作为训练集计算哈希函数,并利用训练好的哈希函数将特征点的浮点型特征描述符映射至汉明空间,实现特征点对间相似度的快速计算,进一步通过剔除误匹配点获得精确匹配结果。同时,训练集中的特征描述符的种类可根据星载平台载荷的传感器类型、影像分辨率及目标区域影像的地貌类型进行灵活选择,使得本文方法具备良好的可重构性;计算汉明距离衡量特征点对间的相似度,提高本文方法在星载平台轻量化处理的应用能力。选取不同时刻资源三号卫星影像及高分七号卫星影像进行匹配对比实验,本文方法轻量化处理后的LW-SIFT方法相较于经典SIFT方法,在耗时方面减少50.12%,且增加正确匹配点数达20.28%。实验结果表明,本文方法能够显著提升影像匹配的精确度及时效性,有较大的应用潜力,能够为星载平台应用提供有力支撑。

关键词: 星载平台快速匹配, 高空间分辨率遥感影像, 高分七号, 资源三号, 影像特征, 描述符轻量化, 哈希映射, 二值型特征描述符

Abstract:

To solve the existing problem that traditional matching methods are difficult to achieve fast matching on the spaceborne platform under the condition of limited storage and computing power. So, this paper proposes a fast image matching algorithm based on spaceborne platform. The feature descriptors of the same-origin satellite images are used as the training set to calculate the hash function, and the floating-point feature descriptors are mapped to the Hamming space to calculate the similarity. The mismatching elimination method is used to get a set of matching points with higher matching accuracy. And the types of feature descriptors can be flexibly adjusted according to the sensor type of the spaceborne platform load, image resolution, and landform type of the target area image, which makes the proposed method have good reconfigurability. The hamming distance is calculated to measure the similarity between feature point pairs and improve the application ability of the method in the light-weight processing of the spaceborne platform. The matching experiments of different geomorphological features using the different orbital imageries from ZY-3 and GF-7 satellites are completed. The experiments prove that compared with the SIFT, the LW-SIFT (Light Weight Scale-Invariant Feature Transform) proposed in the paper can achieve satellite remote sensing image matching more efficiently, reducing time consumption by 50.12% and increasing the number of correct matching points by 20.28 %. This proposed method can significantly improve the accuracy and timeliness of matching and has a large application potential for spaceborne platform applications.

Key words: fast matching on the spaceborne platform, high-resolution satellite images, GF-7 satellite, ZY-3 satellite, image feature, lightweight descriptors, hash algorithm, binary feature descriptors

岳振宇, 范大昭, 董杨, 纪松, 李东子. 一种星载平台轻量化快速影像匹配方法[J]. 地球信息科学学报, 2022, 24(5): 925-939.DOI:10.12082/dqxxkx.2022.210529

YUE Zhenyu, FAN Dazhao, DONG Yang, JI Song, LI Dongzi. A Generation Method of Spaceborne Lightweight and Fast Matching[J]. Journal of Geo-information Science, 2022, 24(5): 925-939.DOI:10.12082/dqxxkx.2022.210529

图/表 12

图1

图2

图3

表1

图4

表2

图5

图6

图7

表3

图8

表4

参考文献 21

[1]	王密, 杨芳. 智能遥感卫星与遥感影像实时服务[J]. 测绘学报, 2019, 48(12):1586-1594.
	[ Wang M, Yang F. Intelligent remote sensing satellite and remote sensing real-time service[J]. Acta Geodaetica et Cartographica Sinica, 2019, 48(12):1586-1594. ] DOI: 10.11947/j.AGCS.2019.20190454 doi: 10.11947/j.AGCS.2019.20190454
[2]	高昆, 刘迎辉, 倪国强, 等. 光学遥感图像星上实时处理技术的研究[J]. 航天返回与遥感, 2008, 29(1):50-54.
	[ Gao K, Liu Y H, Ni G Q, et al. Study on On-board Real-time Image Processing Technology of Optical Remote Sensing[J]. Spacecraft Recovers & Remote Sensing, 2008, 29(1):50-54.(in Chinese). ] DOI: 10.3969/j.issn.1009-8518.2008.01.010 doi: 10.3969/j.issn.1009-8518.2008.01.010
[3]	杨芳, 刘思远, 赵键, 等. 新型智能遥感卫星技术展望[J]. 航天器工程, 2017, 26(5):74-81.
	[ Yang F, Liu S Y, Zhao J, et al. Technology prospective of intelligent remote sensing satellite[J]. Spacecraft Engineering, 2017, 26(5):74-81.(in Chinese). ] DOI: 10.3969/j.issn.1673-8748.2017.05.012 doi: 10.3969/j.issn.1673-8748.2017.05.012
[4]	李德仁, 王密, 沈欣, 等. 从对地观测卫星到对地观测脑[J]. 武汉大学学报·信息科学版, 2017, 42(2):143-149.
	[ Li D R, Wang M, Shen X, et al. From earth observation satellite to earth observation brain[J]. Geomatics and Information Science of Wuhan University, 2017, 42(2):143-149. ] DOI: 10.13203/j.whugis20160526 doi: 10.13203/j.whugis20160526
[5]	Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2):91-110. DOI: 10.1023/B%3AVISI.0000029664.99615.94 doi: 10.1023/B%3AVISI.0000029664.99615.94
[6]	Bay H, Tuytelaars T, Van Gool L. SURF: Speeded up robust features [C]. European Conference on Computer Vision, 2006:404-417. DOI: 10.1007/11744023_32 doi: 10.1007/11744023_32
[7]	许允喜, 陈方. 特征联合和旋转不变空间分割联合的局部图像描述符[J]. 自动化学报, 2016, 42(4):617-630.
	[ Xu Y X, Chen F. Recent advances in local image descriptor[J]. Journal of Image and Graphics, 2015, 20(9):1133-1150. ] DOI: 10.16383/j.aas.2016.c150206 doi: 10.16383/j.aas.2016.c150206
[8]	范大昭, 董杨, 张永生. 卫星影像匹配的深度卷积神经网络方法[J]. 测绘学报, 2018, 47(6):844-853.
	[ Fan D Z, Dong Y, Zhang Y S, et al. Satellite image matching method based on deep convolution neural network[J]. Acta Geodaetica et Cartographica Sinica, 2018, 47(6):844-853. ] DOI: 10.11947/j.AGCS.2018.20170627 doi: 10.11947/j.AGCS.2018.20170627
[9]	董杨. 多视角视频时空四维重建理论与方法研究[D]. 郑州:战略支援部队信息工程大学, 2020.
	[ Dong Y. Spatio-Temporally Coherent 4D Reconstruction from Multiple View Video[D]. Zhengzhou, Information Engineering University, 2020. ] DOI: 10.27188/d.cnki.gzjxu.2020.000169 doi: 10.27188/d.cnki.gzjxu.2020.000169
[10]	Ke Y, Sukthankar R. PCA-SIFT: A more distinctive representation for local image descriptors[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2004:II-II. DOI: 10.1109/cvpr.2004.1315206 doi: 10.1109/cvpr.2004.1315206
[11]	Leutenegger S, Chli M, Siegwart R Y. BRISK: Binary robust invariant scalable keypoints [C]. IEEE International Conference on Computer Vision, 2011. DOI: 10.1109/iccv.2011.6126542 doi: 10.1109/iccv.2011.6126542
[12]	Rublee E, Rabaud V, Konolige K, et al. ORB: An efficient Alternative to SIFT or SURF [C]. IEEE International Conference on Computer Vision, 2011. DOI: 10.1109/iccv.2011.6126544 doi: 10.1109/iccv.2011.6126544
[13]	李恩, 温志盛, 章育林, 等. 基于OpenCV与均值哈希算法的人脸相似识别系统[J]. 工业设计, 2016, 121(8):132-133.
	[ Li E, Wen Z S, Zhang Y L, et al. Face recognition system based on OpenCV and mean hash algorithm[J]. Industrial Design, 2016, 121(8):132-133. ]
[14]	方杨. 基于深度学习的鲁棒图像哈希检索方法研究[D]. 武汉:华中科技大学, 2019.
	[ Fang Y. Research on robust hash of image retrieval based on deep learning[D]. Wuhan, Huazhong University of Science & Technology, 2019. ] DOI: 10.27157/d.cnki.ghzku.2019.003595 doi: 10.27157/d.cnki.ghzku.2019.003595
[15]	陈苏婷, 郭子烨, 张艳艳. 基于局部哈希学习的大面阵CCD航拍图像匹配方法[J]. 应用光学, 2019, 40(2):259-264.
	[ Chen S T, Guo Z Y, Zhang Y Y. Matching method of large array CCD aerial images based on local hashing learning[J]. Journal of Applied Optics, 2019, 40(2):259-264. ] DOI: 10.5768/JAO201940.0202003 doi: 10.5768/JAO201940.0202003
[16]	冯瑞. 基于HSI哈希学习的航拍图像匹配算法研究[D]. 南京:南京信息工程大学, 2019.
	[ Feng R. Aerial image matching method based on HSI hash learning[D]. Nanjing: Nanjing Information Engineering University, 2019. ] DOI: 10.27248/d.cnki.gnjqc.2019.000329 doi: 10.27248/d.cnki.gnjqc.2019.000329
[17]	霍门婕. 在轨影像控制点的轻量化设计与应用[D]. 郑州:战略支援部队信息工程大学, 2020.
	[ Huo M J. Lightweight design and application of in-orbit image control points[D]. Zhengzhou: Information Engineering University, 2020. ] DOI: 10.27188/d.cnki.gzjxu.2020.000171 doi: 10.27188/d.cnki.gzjxu.2020.000171
[18]	Strecha C, Bronstein A, Bronstein M, et al. LDAHash: Improved matching with smaller descriptors[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(1):66-78. DOI: 10.1109/tpami.2011.103 doi: 10.1109/TPAMI.2011.103 pmid: 21576750
[19]	Blei D M, Ng A Y, Jordan M I. Latent dirichlet allocation[J]. The Annals of Applied Statistics, 2001.
[20]	Fischler M A, Bolles R C. Random Sample Consensus: A paradigm for model fitting with applications to image analysis and automated cartography-sciencedirect[J]. Readings in Computer Vision, 1987:726-740.
[21]	李德仁. 我国第一颗民用三线阵立体测图卫星——资源三号测绘卫星[J]. 测绘学报, 2012, 41(3):317-322.
	[ Li D R. China's first civilian three-line-array stereo mapping satellite: ZY-3[J]. Acta Geodaetica et Cartographica Sinica, 2012, 41(3):317-322. ]

	实验1		实验2
	左影像	右影像	左影像	右影像	左影像	右影像
卫星名称	ZY3-1	ZY3-1	GF7	GF7	GF7	GF7
传感器	NAD	NAD	BWD	BWD	FWD	FWD
采集时间	2019-05-22	2020-05-10	2020-12-15	2020-12-15	2020-12-30	2020-12-30
分辨率/m	2.10	2.10	0.65	0.65	0.80	0.80

序号	特征检测方法	描述符生成方法	描述符维度	描述符存储类型		描述符匹配方法
1	SIFT	SIFT	128	float	4096	ANN
2	LW-SIFT	SIFT-Hamming	128	unsigned char	128	BF-Hamming
3	SURF	SURF	64	float	2048	ANN
4	LW-SURF	SURF-Hamming	64	unsigned char	64	BF-Hamming
5	BRISK	BRISK	64	unsigned char	64	BF
6	ORB	ORB	32	unsigned char	32	BF-Hamming

匹配方法	地貌类型	影像数量	检测特征点总数/个	RANSAC后正确匹配总数/个	匹配总耗时/s	平均特征点增加比率/%	平均耗时减少比率/%
SIFT	农田	27	1 179 572	26 503	433.544	-	-
	山地	46	2 253 819	94 751	853.594	-	-
	居民地	13	988 022	46 923	288.170	-	-
	城镇	35	2 108 630	60 043	821.901	-	-
LW-SIFT	农田	27	925 733	31 153	178.002	18.87	57.26
	山地	46	1 766 759	114 827	367.511	21.19	54.18
	居民地	13	768 160	56 137	231.192	19.64	39.50
	城镇	35	1 641 941	72 876	380.483	21.37	49.56

影像类型	匹配方法	检测特征点总数/个	RANSAC优化后正确匹配总数/个	匹配总耗时/s	平均特征点增加比率/%	平均耗时减少比率/%
G7-BWD (80-23)	SIFT	1 286 691	23 827	534.625	-	-
	LW-SIFT	1 015 306	28 040	299.209	16.59	32.36
	SURF	1 626 460	40 505	600.890	-	-
	LW-SURF	942 109	8 667	224.770	-75.80	60.45
	BRISK	1 202 657	28 202	182.361	-	-
	ORB	4 186 775	130 829	1 143.461	-	-
G7-FWD (36-10)	SIFT	304 796	3 129	145.891	-	-
	LW-SIFT	240 096	3 648	98.684	17.68	44.03
	SURF	639 063	6 566	222.127	-	-
	LW-SURF	376 989	1 589	87.859	-78.60	62.59
	BRISK	267 828	6 631	51.944	-	-
	ORB	990 204	25 187	218.379	-	-