![Contact Dr. Oliver Lang Senior Application Development Manager Development & Engineering | Infoterra GmbH GEO-Information Services Astrium GmbH - Services Claude-Dornier-Str. | 88090 Immenstaad | Germany Tel +49 7545 8 5520 | Fax +49 7545 8 1337 | Mob +49 151 1822 0827 [email_address] | www.infoterra.de](https://image.slidesharecdn.com/integration-of-colorized-single-pol-sar-data-into-optical-image-mosaics-110727153339-phpapp01/85/Integration-of-Colorized-Single-Pol-SAR-Data-Into-Optical-Image-Mosaics-ppt-15-320.jpg)
Integration of Colorized Single-Pol SAR Data Into Optical Image Mosaics.ppt
- 1. Integration of Colorized Single-Pol SAR Data Into Optical Image Mosaics Oliver Lang Parivash Lumsdon Astrium GEO-Information Services IGARSS 2011, Vancouver
- 2. Motivation Thematic mapping in Cloud Belt using single-pol SAR Cost effective approach: single coverage, full resolution + swath width Mosaic electro-optical image mosaics with seamless SAR mosaics, colorized in meaningful way Commercial TerraSAR-X data distribution by Astrium: Colored quick looks come with TSX data since 2010 BUT: varying colors, not intuitive Development of new add-on product: Color Composite
- 3. Single-Pol SAR Colorization Known: Basic „classification“ based on Speckle variations Coeff. of Variation as measure for local speckle noise Link to main surface types: Large CoV: heterogenious (urban) Small CoV: homogenious (water, grassland) New: combination of multiple texture filters Colorization according to reference image S. Kuntz and F. Siegert, “Monitoring of deforestation and land use in Indonesia with multitemporal ERS data.” International Journal of Remote Sensing 20: 2835-2853, 1999 M. Thiel., T. Esch, and S. Dech, “Object-oriented detection of settlement areas from TerraSAR-X data” Proceedings of the EARSeL Joint Workshop: Remote Sensing: New Challenges of high resolution. (Eds.,Carsten Jürgens), 2008 STD mean
- 4. General Approach Apply multiscale texture filters Classification based on filter layers Colorization of „classes“ with given LUTs
- 5. Generation of filter layers Derivation of multi-scale texture components Mean Standard Deviation Variance Skewness Coeff of Variation Spectral high-pass Noise components: Multiplicative Noise S : apply Gaussian filter Additive Noise N = apply directional Lee filtered
- 6. Classification Hierarchical unsupervised classification based on filter layers Min-distance based on empirical thresholds Backscatter & speckle characteristics allows reliable separaton of (calm) Water / Urban Third class is separated into hetero- and honogenious sub-class (e.g. Forest / Grassland ) Decider: local Variance Histogram value Urban Forest Water 0 50 100 150 200 250 mean STD
- 7. Selection of Colors 2 methods: Predefined standard color tables (optical) reference image Manual or automatic selection of samples for each class Derivation of Hue values from samples and quantization of colors to a desired number of colors 4 LUTs HSV RGB Transformation Example: selection of sample areas Background image: Google Earth mean STD hue Saturation
- 8. 2 Examples Overlay: Spot 4 and TerraSAR-X Stripmap Overlay: TerraSAR-X Spotlight in Google Earth
- 9. Example: Cameroon TerraSAR-X: Date: 29 Jul 2010 StripMap, 3 m res HH polarization SPOT4 : date: 8 Jan 2011 20 m resolution, Layers 4, 1, 2 10 km
- 10. Color tables derived from overlapping optical scene Nr. of quantized colors: 16 Example: Cameroon TerraSAR-X: Date: 29 Jul 2010 StripMap, 3 m res HH polarization SPOT4 : date: 8 Jan 201 20 m resolution, Layers 4, 1, 2 Water Agriculture Forest Urban 10 km
- 11. Example: Germany Quantization: 256 colors / class Germany: TerraSAR-X HS urban forest agriculture water
- 12. Example: Germany Background image: Google Earth Germany: TerraSAR-X HS urban forest agriculture water
- 13. Discussion Sensor and SAR-mode independent qualitative approach Supports thematic mapping as additional information layer, e.g. in cloud belt Intuitive visualization and interactive interpretation SAR specific backscatter characteristics remain Inherently, differences regarding surface representation between optical and SAR remain Further improvements expected by optimized classification procedure & automatic LUT derivation 0% 100% Clouds
- 14. Thank You Astana, Kasachstan TerraSAR-X StripMap Color SAR
- 15. Contact Dr. Oliver Lang Senior Application Development Manager Development & Engineering | Infoterra GmbH GEO-Information Services Astrium GmbH - Services Claude-Dornier-Str. | 88090 Immenstaad | Germany Tel +49 7545 8 5520 | Fax +49 7545 8 1337 | Mob +49 151 1822 0827 [email_address] | www.infoterra.de
Editor's Notes
- #13: SatDSig Hinweis: Die Bilddimensionen betragen: 14550m (Ost-West), 8790m (Nord-Süd) beziehungsweise 28187 Pixel x 17176 Pixel dies entspricht bei einer vorgegebenen Bildschrimauflösung eine Bildgröße von 31,59 cm x 21,17cm In PPT: 786 x 509 Pix resolution: ca. 18m