![Kmeans clustering for grids
Processing → Toolbox → SAGA (2.3.1) [geoalgorithms] → Image analysis
→ Kmeans clustering for grids](https://image.slidesharecdn.com/op04-220506072354-d1c856ae/85/Data-Processing-Using-DIWATA-1-Microsatellite-Imagery-for-Disaster-Monitoring-38-320.jpg)
Data Processing Using DIWATA-1 Microsatellite Imagery for Disaster Monitoring
- 1. Center for Research and Application for Satellite Remote Sensing Yamaguchi University Data Processing Using DIWATA-1 Microsatellite Imagery for Disaster Monitoring
- 2. Diwata-1 • Diwata-1 is the Philipines’ first micro-satellite with a baseline design of the 50 kilogram. • The payload subsystem includes a high-precision telescope (HPT) with 3 meter spatial resolution and bands R/G/B/NIR. • The second payload is a space-borne multi-spectral imager (SMI) with liquid crystal tunable filter (LCTF), which has an 80 meter spatial resolution. • The LCTF allows for image capture to be taken at 10 nanometer increments from 420nm – 650nm for the visible range and 700nm – 1050nm for the near infrared region. • With a spatial resolution of 7 kilometers, the wide field camera can used to complement current weather forcasting capabilities. This sensor can potentially image cloud formations and typhoons. phl-microsat.upd.edu.ph/diwata1 NASA – space.skyrocket.de
- 3. Data Download Link The following is the link for data downloading: http://data.phl-microsat.xyz
- 4. Creating a New Account Fill necessary information related to: (1) Account information; (2) Personal information;
- 6. (4) End User Agreement
- 9. Criteria Selection for Data Searching Criteria selection Satellite: Diwata Payload: SMI; MFC Data of Acquisition: All Available data location will be shown in color with information of the data number in the period of selected date of acquisition
- 10. Data Availability • Available data for DIWATA-1 imageries (2016 – 2018)
- 11. Selected area in 2016. The graph below the image indicates the number of available data. • Area of interest can be selected by using “Draw a rectangle” or “Draw a marker” option. • The available data will be provided on the right hand-side. • Intended data can be checked from the “view image details” option. • Data can be downloaded directly by using “Download image” option or collected first by choosing “add image to download cart” option. Area Selection
- 12. Visible band images of the downloaded raw images Visible - 550 nm Visible - 710 nm Visible - 440 nm Blue band 450 - 495 nm Green band 495 - 570 nm Red band 620 - 750 nm
- 13. Near-infrared (NIR) band images of the downloaded raw data NIR - 750 nm NIR - 760 nm NIR - 870 nm
- 14. • Vegetation index calculation using data sets from different acquisition date Location: Sablayan, Occidental Mindoro, Philippines Image acquisition date: 2016/04/03 and 2017/04/21 • Post-disaster analysis of flood event on March 12th, 2017 Location: Lamitan City, Basilan Province, Philippine Image acquisition date: 2017/07/01 Image processing using DIWATA-1
- 15. Vegetation index calculation using data sets from different acquisition dates • Location: Sablayan, Occidental Mindoro, Philippines
- 16. Data collection according to the location and availability (1) 2016/11/10 → 1st Data set (2) 2017/04/21 → 2nd Data set
- 17. Band selection for Visible (V) and Near-Infrared (N) (1) 2016/11/10 → 1st Data set Red (V) NIR (N) Red (V) NIR (N) Red (V) NIR (N) Red (V) NIR (N) NIR (N) NIR (N) Red (V) Red (V) • Band selection was conducted according to the location and visibility of the image. • Selected bands are one Visible (Red – 680 mm) band and one NIR (N – 740 mm) band. Selected bands for NDVI
- 18. Band selection for Visible (V) and Near-Infrared (N) (2) 2017/04/21 → 2nd Data set NIR (V) NIR (N) NIR (N) NIR (N) NIR (N) Green (V) Green (V) Blue (V) Red (V) Blue (V) Selected bands for NDVI • Visible bands selected from the data set are Red – 710 mm, Green – 530 mm, and Blue – 480 mm. • Near-Infrared band selected from the data set is NIR – 740 mm.
- 19. Build virtual raster from selected bands • Raster → Miscellaneous → Build Virtual Raster (Catalog) • Virtual Raster: Band 1 : V440 (Blue band) Band 2 : V530 (Green band) Band 3 : V710 (Red band) Band 4 : N740 (NIR band)
- 20. True-color composite • Right click on the virtual raster layer → Properties → Style R: V710; G: V530; B: V480 2017/04/21 – True Color
- 21. False color composite • Right click on the virtual raster layer → Properties → Style → R: N740; G: V710; B: V530 2017/04/21 – False Color
- 22. Checking geographical boundary with available reference map *Georectified should be conducted in case the image is not in accordance with the reference map. • “Layers Panel” → 🗹 Philippine administration shapefile (PHL_adm0) 🗹 Virtual raster layer of the image
- 23. NDVI calculation N (NIR) band – V (R) band / N (NIR) band + V (R) band • Raster → Raster Calculator 1st Data set (2016/11/10) NDVI = (ρ740 – ρ680) / (ρ740 + ρ680) 2nd Data set (2017/04/11) NDVI = (ρ740 – ρ680) / (ρ740 + ρ680)
- 24. Changing NDVI color • Properties → Style ”Color” is changed into “bgyr” color type
- 25. NDVI image results 1st Data set (2016/11/10) NDVI = (ρ740 – ρ680) / (ρ740 + ρ680) 2nd Data set (2017/04/11) NDVI = (ρ740 – ρ680) / (ρ740 + ρ680) 1 -1 -1 1 -1 1 -1 1
- 26. Post-disaster analysis of flood event Location: Lamitan City, Basilan Province, Philippine As of March 12th, 2017, over 11,000 people in Lamitan City (Basilan Province) and Tambulig municipality (Zambonga del Sur province) are displaced by flooding. The floods damaged 127 houses, mostly in Lamitan City (OCHA, 2017).
- 27. Data collection according to the location and availability Post-disaster: 2017/07/01
- 28. Band selection for Visible (V) and Near-Infrared (N) NIR (N) NIR (N) NIR (N) NIR (N) Red (V) Red (V) Blue (V) Red (V) Red (V) Red (V) Green (V) Blue (V) Green (V) Visible bands: Red – 690 mm; Green – 550 mm; Blue – 490 mm Near-Infrared band: NIR – 760 mm Selected bands for analysis
- 29. Build virtual raster from selected bands • Raster → Miscellaneous → Build Virtual Raster (Catalog) Virtual Raster: Band 1 : V490 (Blue band) Band 2 : V550 (Green band) Band 3 : V690 (Red band) Band 4 : N760 (NIR band)
- 30. True-color composite • Right click on the virtual raster layer → Properties → Style R: V690; G: V550; B: V490 2017/07/01 – True Color
- 31. False color composite • Right click on the virtual raster layer → Properties → Style → R: N760; G: V690; B: V550 2017/07/01 – False Color
- 32. Checking geographical boundary with available reference map *Georectified should be conducted in case the image is not in accordance with the reference map. • “Layers Panel” → 🗹 Philippine administration shapefile (PHL_adm0) 🗹 Virtual raster layer of the image
- 33. NDWI calculation V (G) band – N (NIR) band / V (G) band + N (NIR) band • Raster → Raster Calculator • NDWI = (ρ550 – ρ760) / (ρ550 + ρ760 1 -1
- 34. Changing NDVI color 1 -1 • Properties → Style ”Color” is changed into “bgyr” color type
- 35. NDVI calculation N (NIR) band – V (R) band / N (NIR) band + V (R) band • Raster → Raster Calculator • NDVI = (ρ760 – ρ690) / (ρ760 + ρ690) 1 -1
- 36. Changing NDVI color -1 1 Properties → Style ”Color” is changed into “bgyr” color type
- 37. Histogram of the NDVI image result as reference for image segmentation • Right click on the NDVI image’s raster layer → Properties → Histogram Histogram of DIWATA-1 Post-disaster NDVI image result 2017/07/01
- 38. Kmeans clustering for grids Processing → Toolbox → SAGA (2.3.1) [geoalgorithms] → Image analysis → Kmeans clustering for grids
- 39. Sieving the clustered image result • Raster → Analysis → Sieve
- 40. Polygonize raster layer to vector layer • Raster → Conversion → Polygonize (Raster to vector)
- 41. Changing properties of vector layer • Right click on the sieved cluster’s vector layer → Properties → Style
- 42. Save vector layer as shapefile • Right click on the sieved cluster’s vector layer → Save vector layer as
- 43. Landsat-8 data analysis for the flood event • Pre-Disaster (2016/04/03) • Post-Disaster (2017/04/22) True-color composite RGB – 432 True-color composite RGB – 432 False-color composite RGB – 543 False-color composite RGB – 543 Obtained changes between pre- and post-disaster images
- 44. NDVI comparison between pre- & post-disaster of Landsat-8 and post-disaster of DIWATA-1 -1 1 -1 1 2016/04/03 – Pre-Disaster 2017/04/22 – Post-Disaster 2017/07/01 – Post-Disaster Landsat-8 Landsat-8 -1 1 Diwata-1 Flood line
- 45. Flood line on pre- and post-disaster images Post-Disaster DIWATA-1 (2017//07/01) Pre-Disaster Landsat-8 (2016/04/03) Flood line • Overlaying raster and vector layers by activating the layers → Check the layers in the “Layers Panel” • Active layers: 🗹 True-color composite raster layer 🗹 Cluster polygon shapefile of flood line
- 46. Map preparation Project → New Print Composer Add map → Add legend, draw grid, draw north arrow, and add label
- 47. Landsat-8 True-color composite RGB 4-3-2 Pre-Disaster (2016/04/03) • Composer → Export as image Export map as image DIWATA-1 False-color composite RGB 4-3-2 Pre-Disaster (2016/04/03)