SAR Change Detection Algorithm Based on Space-Frequency Dual-Domain Filtering

Yuqing Wu; Qing Xu; Jingzhen Ma; Bowei Wen; Xinming Zhu; Tianming Zhao

doi:10.3788/AOS221834

Acta Optica Sinica, Volume. 43, Issue 12, 1228009(2023)

SAR Change Detection Algorithm Based on Space-Frequency Dual-Domain Filtering

Yuqing Wu... Qing Xu*, Jingzhen Ma, Bowei Wen, Xinming Zhu and Tianming Zhao |Show fewer author(s)

Author Affiliations

Institute of Geospatial Information, Information Engineering University, Zhengzhou 450001, Henan, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(30)

Fig. 1. Process of SAR image change detection method based on dual-domain filtering

Download full size

Fig. 2. Laplace airspace fusion scheme

Download full size

Fig. 3. Laplace pyramid generation

Download full size

Fig. 4. Flowchart of frequency domain filtering

Download full size

Fig. 5. SAR images of Bern area. (a) Image taken in April 1999; (b) image taken in May 1999; (c) change reference map

Download full size

Fig. 6. SAR images of Ottawa area. (a) Image taken in May 1997; (b) image taken in August 1997; (c) change reference map

Download full size

Fig. 7. SAR images of San Francisco area. (a) Image taken in August 2003; (b) image taken in May 2004; (c) change reference map

Download full size

Fig. 8. SAR images of Yellow River area. (a) Image taken in June 2008; (b) image taken in June 2009; (c) change reference map

Download full size

Fig. 9. Results of change detection in Bern dataset using K-means clustering

Download full size

Fig. 10. Results of change detection in Ottawa dataset using K-means clustering

Download full size

Fig. 11. Results of change detection in San Francisco dataset using K-means clustering

Download full size

Fig. 12. Results of change detection in Yellow River dataset using K-means clustering

Download full size

Fig. 13. Results of change detection in Bern dataset using FCM clustering

Download full size

Fig. 14. Results of change detection in Ottawa dataset using FCM clustering

Download full size

Fig. 15. Results of change detection in San Francisco dataset using FCM clustering

Download full size

Fig. 16. Results of change detection in Yellow River dataset using FCM clustering

Download full size

Fig. 17. Kappa coefficients of Bern dataset in different clustering methods. (a) K-means clustering method; (b) FCM clustering method

Download full size

Fig. 18. Comparison of results obtained by different algorithms

Download full size

Fig. 19. Relationship between PCC and cutoff frequency. (a) K-means clustering method; (b) FCM clustering method

Download full size

Table 1. Metrics of Bern dataset Based on K-means clustering
View table
Table 1. Metrics of Bern dataset Based on K-means clustering
Method FP FN OE PCC /% Kappa /%
Lr 360 326 686 99.24 70.34
Frequency Domain 172 201 373 99.59 83.44
Spatial Domain 152 305 457 99.50 78.56
L-S 204 204 408 99.55 82.11
Dual Domain 128 176 304 99.66 86.39

Table 2. Metrics of Ottawa dataset based on K-means clustering
View table
Table 2. Metrics of Ottawa dataset based on K-means clustering
Method FP FN OE PCC /% Kappa /%
Lr 2086 2741 4827 95.24 81.84
Frequency Domain 712 2317 3029 97.02 88.32
Spatial Domain 1113 2576 3689 96.37 85.82
L-S 1033 2207 3240 96.81 87.64
Dual Domain 418 1860 2278 97.76 91.25

Table 3. Metrics of San Francisco dataset based on K-means clustering
View table
Table 3. Metrics of San Francisco dataset based on K-means clustering
Method FP FN OE PCC /% Kappa /%
Lr 1782 408 2190 96.66 77.83
Frequency Domain 1474 398 1872 97.14 80.55
Spatial Domain 1469 411 1880 97.13 80.43
L-S 935 761 1696 97.41 80.83
Dual Domain 395 662 1057 98.39 87.52

Table 4. Metrics of Yellow River dataset based on K-means clustering
View table
Table 4. Metrics of Yellow River dataset based on K-means clustering
Method FP FN OE PCC /% Kappa /%
Lr 11144 5466 16610 77.64 35.29
Frequency Domain 8444 3798 12242 83.52 50.98
Spatial Domain 10222 4338 14560 80.40 43.48
L-S 6566 5986 12552 83.10 43.90
Dual Domain 1984 2655 4639 93.75 78.50

Table 5. Metrics of Bern dataset based on FCM Clustering
View table
Table 5. Metrics of Bern dataset based on FCM Clustering
Method FP FN OE PCC /% Kappa /%
Lr 430 295 725 99.20 69.94
Frequency Domain 184 193 377 99.58 83.41
Spatial Domain 152 305 457 99.50 78.56
L-S 234 185 419 99.54 82.00
Dual Domain 131 174 305 99.66 86.38

Table 6. Metrics of Ottawa dataset based on FCM clustering
View table
Table 6. Metrics of Ottawa dataset based on FCM clustering
Method FP FN OE PCC /% Kappa /%
Lr 2106 2723 4829 95.24 81.85
Frequency Domain 661 2420 3081 96.96 88.07
Spatial Domain 1113 2576 3689 96.37 85.82
L-S 1010 2245 3255 96.79 87.57
Dual Domain 361 1996 2357 97.68 90.90

Table 7. Metrics of San Francisco dataset based on FCM clustering
View table
Table 7. Metrics of San Francisco dataset based on FCM clustering
Method FP FN OE PCC /% Kappa /%
Lr 1745 409 2154 96.71 78.12
Frequency Domain 1474 398 1874 97.14 80.55
Spatial Domain 1469 411 1880 97.13 80.43
L-S 2186 439 2625 95.99 74.26
Dual Domain 928 390 1318 97.99 85.61

Table 8. Metrics of Yellow River dataset based on FCM clustering
View table
Table 8. Metrics of Yellow River dataset based on FCM clustering
Method FP FN OE PCC /% Kappa /%
Lr 12642 5091 17733 76.12 33.90
Frequency Domain 9942 3466 13428 81.92 48.60
Spatial Domain 10222 4338 14560 80.40 43.48
L-S 13872 3890 17702 76.09 37.41
Dual Domain 2469 3138 5607 92.45 74.02

Table 9. Kappa coefficient result

View table

Table 9. Kappa coefficient result

Method	Bern	Ottawa	San Francisco	Yellow River
K-means clustering	70.34	81.84	77.83	35.29
FCM clustering	69.94	81.85	78.12	34.29
PCAKM	85.75	88.17	88.80	78.37
FLICM	85.82	89.47	88.62	71.18
PCANet	75.37	93.11	91.22	82.43
Dual Domain-K	86.39	91.25	87.52	78.50
Dual Domain-F	86.38	90.90	85.61	74.02

Table 10. Kappa coefficients for K-means clustering at different fusion coefficients
View table
Table 10. Kappa coefficients for K-means clustering at different fusion coefficients
Fusion coefficient Bern Ottawa San Francisco Yellow River
$q = 0.5$ and $w = 0.5$ 83.39 91.25 87.52 74.27
$q = 0.8$ and $w = 0.2$ 85.82 90.09 84.53 68.11

Table 11. Kappa coefficients for FCM clustering at different fusion coefficients
View table
Table 11. Kappa coefficients for FCM clustering at different fusion coefficients
Fusion coefficient Bern Ottawa San Francisco Yellow River
$q = 0.5$ and $w = 0.5$ 86.38 90.90 85.61 74.02
$q = 0.8$ and $w = 0.2$ 85.86 89.78 83.25 65.95

Tools

Get Citation

Copy Citation Text

Yuqing Wu, Qing Xu, Jingzhen Ma, Bowei Wen, Xinming Zhu, Tianming Zhao. SAR Change Detection Algorithm Based on Space-Frequency Dual-Domain Filtering[J]. Acta Optica Sinica, 2023, 43(12): 1228009

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites