Title Page
Contents
요약문 7
Abstract 9
1. Introduction 13
1.1. Research Background 13
1.2. Research Purpose 15
2. Literature Review 18
3. Research Data 22
3.1. Water Database Collection 22
3.2. SAR (Sentinel-1) and Optical Imagery (Sentinel-2) 24
3.3. Ground Truth Data 29
4. Study Area and Methods 31
4.1. Study area and Data 31
4.2. Water Extraction Method 40
4.3. Surface Water Change Analysis 46
4.4. Validation and Accuracy 51
4.4.1. Visual Comparation with GSW (Global Surface Water) 51
4.4.2. Accuracy Assessment 52
5. Result and Discussion 54
5.1. Analysis Results 54
5.2. Discussion 65
6. Conclusion 66
References 68
Table 1. Details of data collection of digital maps for water DB. 22
Table 2. Sentinel-1 data characteristics. 25
Table 3. Sentinel-2 data characteristics. 28
Table 4. Detail data for ground truth data making. 31
Table 5. The data sets for 2022s, which include SAR and Sentinel-2 images. 39
Table 6. The data sets for 2021s, which include SAR and Sentinel-2 images. 39
Table 7. The data sets for 2020s, which include SAR and Sentinel-2 images. 39
Table 8. The data sets for 2019s, which include SAR and Sentinel-2 images. 40
Table 9. The data sets for 2018s, which include SAR and Sentinel-2 images. 40
Table 10. Detail of SAR-C data which used for first model in South Korea sample. 50
Table 11. The result of the surface water total area changed from the first model. 51
Table 12. The overall total area of surface water area in North Korea from 2018 to 2022. 62
Table 13. All accuracy result from F₁-score accuracy assessment for five selected samples. 63
Table 14. Accuracy result from F₁-score for five selected samples of reference data from GSW. 64
Figure 1. Workflow process of making a water DB from digital map and land cover map 23
Figure 2. Appearance of water DB and detail of water properties with a V World Map as a base map. a. Water DB raster type. b. Water DB polygon... 24
Figure 3. The workflow for the SAR images pre-processing in SNAP and the SNAP tools user interface. 26
Figure 4. The pre-processing process in SNAP tools. 27
Figure 5. Flow process ground truth data making. 30
Figure 6. The study area of North Korea region with covered by the sample images using the Sentinel-1 (SAR) and Sentinel-2 L2A data. 32
Figure 7. The study area with the samples data from 2018. 33
Figure 8. The study area with the samples data from 2019. 34
Figure 9. The study area with the samples data from 2020. 35
Figure 10. The study area with the samples data from 2021. 36
Figure 11. The study area with the samples data from 2022. 37
Figure 12. Automated water detection workflow process. 42
Figure 13. Detail process of image segmentation and binary mask. 44
Figure 14. The overall workflow processes. 45
Figure 15. Surface water change analysis workflow. 48
Figure 16. Additional experiment flow for the surface water changed validity. 49
Figure 17. Three sample location of first model which represents big river in South Korea. 50
Figure 18. Global Surface Water (GSW) mask reference data. a. Global explorer platform. b. GSW mask. 52
Figure 19. The comparation result from this approach water mask with GSW reference data. 55
Figure 20. Sample results from 2018 data period 56
Figure 21. Sample results from 2019 data period. 57
Figure 22. Sample results from 2020 data period. 58
Figure 23. Sample results from 2021 data period. 59
Figure 24. Sample results from 2022 data period. 60
Figure 25. The water mask result from the automated water detection. The black area represents non-water area, and the white area represents the water area. 62
Figure 26. Graphic show the surface water area from 2018 to 2022 in North Korea is decrease. 63