Acta Optica Sinica, Volume. 44, Issue 6, 0628007(2024)
Phase Congruency Satellite Image Matching Method Based on Anisotropic Filtering
Figures & Tables(22)
Fig. 2. Scale space construction. (a) Gaussian scale space; (b) nonlinear scale space
Fig. 3. Image blocks and maximum moment maps. (a) Original image blocks; (b) maximum moment maps
Fig. 4. Detection of feature points on maximum moment map. (a) Maximum moment map feature points of reference image block; (b) maximum moment map feature points of image block to be matched
Fig. 13. Matching results of experimental data of group A. (a) SIFT algorithm; (b) RIFT algorithm; (c) HAPCG algorithm; (d) proposed algorithm
Fig. 14. Matching results of experimental data of group B. (a) SIFT algorithm; (b) RIFT algorithm; (c) HAPCG algorithm; (d) proposed algorithm
Fig. 15. Matching results of experimental data of group C. (a) SIFT algorithm; (b) RIFT algorithm; (c) HAPCG algorithm; (d) proposed algorithm
Fig. 16. Matching results of experimental data of group D. (a) SIFT algorithm; (b) RIFT algorithm; (c) HAPCG algorithm; (d) proposed algorithm
Fig. 17. Matching results of experimental data of group E. (a) SIFT algorithm; (b) RIFT algorithm; (c) HAPCG algorithm; (d) proposed algorithm
Fig. 18. Matching results of experimental data of group F. (a) SIFT algorithm; (b) RIFT algorithm; (c) HAPCG algorithm; (d) proposed algorithm
Table 1. Experimental data of satellite images
View tableTable 1. Experimental data of satellite images
Dataset Source Year of capture of image Resolution /m Left image Right image Left image Right image Left image Right image A ZY3-01 ZY3-01 2015 2015 3.5 3.5 B ZY3-01 ZY3-01 2015 2015 3.5 3.5 C ZY3-01 ZY3-01 2014 2014 3.5 3.5 D GF-1 ZY3-01 2019 2015 2.0 3.5 E GF-1 ZY3-01 2020 2015 2.0 3.5 F GF-1 ZY3-01 2020 2014 2.0 3.5
Table 2. Matching performance of different matching algorithms
View tableTable 2. Matching performance of different matching algorithms
Dataset SIFT RIFT HAPCG Proposed method NCM RMSE NCM RMSE NCM RMSE NCM RMSE A 944 0.925 1158 1.288 1575 1.181 3084 1.106 B 903 1.009 642 1.313 879 1.113 1230 1.165 C 9 0.871 139 1.371 1096 1.255 1270 1.188 D 9 0.987 5 1.307 40 1.252 136 1.192 E — — 7 1.367 65 1.090 120 1.168 F — — — — 56 1.117 103 1.172
Table 3. Matching performance of different scale space construction methods
View tableTable 3. Matching performance of different scale space construction methods
Dataset Gaussian linearscale space Anisotropic filteringnonlinear scale space NCM RMSE NCM RMSE A 2681 1.207 3084 1.106 B 1101 1.061 1230 1.165 C 1088 1.309 1270 1.188 D 37 1.331 136 1.192 E 60 1.056 120 1.168 F 99 1.024 103 1.172
Table 4. Matching computational efficiency of different CPU cores
View tableTable 4. Matching computational efficiency of different CPU cores
Dataset Time /s Speed-up ratio Single-core CPU Multi-core CPU A 312.784 90.662 3.45 B 279.135 86.688 3.22 C 343.311 90.823 3.78 D 398.073 101.809 3.91 E 380.557 103.694 3.67 F 359.467 100.130 3.59
Tools
Get Citation
Copy Citation Text
Qing Fu, Chen Guo, Wenlang Luo, Shikun Xie. Phase Congruency Satellite Image Matching Method Based on Anisotropic Filtering[J]. Acta Optica Sinica, 2024, 44(6): 0628007
Paper Information
Category: Remote Sensing and Sensors
Received: Nov. 2, 2023
Accepted: Jan. 5, 2024
Published Online: Mar. 19, 2024
The Author Email: Wenlang Luo (9920150045@jgsu.edu.cn)
CSTR:32393.14.AOS231728
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20