SAR Processing And Data Analysis - NASA

National Aeronautics andSpace AdministrationARSETApplied Remote Sensing Traininghttp://arset.gsfc.nasa.gov@NASAARSETSAR Processing and Data AnalysisErika Podestwww.nasa.gov

Learning Objective1. Understand Sentinel Data2. Perform image preprocessing3. Analyze SAR imagery to classify land and waterNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program2

Outline1. Sentinel-1 Background2. Accessing, Opening, and Displaying the Data3. Preprocessing4. AnalysisNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program3

Sentinel-1 Background

Characteristics of SAR Images from Different SatellitesNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program5

Sentinel-1: Modes of Acquisition1. Extra Wide Swath– for monitoring oceans & coasts– 400 km swath, 25 x 40 m spatial resolution2. Strip Mode– special order only– 80 km swath, 5 x 5 m spatial resolution3. Wave Mode– routine collection for the ocean– 20 km swath, 5 x 5 m spatial resolution4. Interferometric Wide Swath– routine collection for land– 250 km swath, 5 x 20 m spatial resolutionNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program6

How to Access Sentinel-1 Images Alaska SAR Facility– http://www.asf.alaska.edu/sentinel/ European Space Agency Portal– ccess-to-sentinel-data/National Aeronautics and Space AdministrationApplied Remote Sensing Training Program7

File Naming Format There are three types of producttypes: SLC, GND, and OCN– SLC: Single Look Complex– GND: Ground Range Detected(You should select this one) Full Resolution (FR), HighResolution (HR), MediumResolution (MR) Resolution is dependent on theamount of multi-lookingperformed– OCN: Level-2 OceanNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program8

Sentinel-1 Toolbox Free and open source software developed by ESA for processing andanalyzing radar images from Sentinel-1 and other satellites Can be accessed through: http://step.esa.int/main/download Includes the following tools:– Calibration– Speckle Noise– Terrain Correction– Mosaic Production– Polarimetry– Interferometry– ClassificationNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program9

Accessing, Opening, and Displaying SARData

Accessing Sentinel-1 Data1. Go to the Alaska Satellite Facility Sentinel Data Portal:http://vertex.daac.asf.alaska.edu/2. Identify your area 92,-60.31,4.52) and dates (Apr 25-29, 2015) of interest3. Identify images of interest (Sentinel-1 A/B)4. Click Search5. Select Granule:S1A IW GRDH 1SDV 20150428T093856 20150428T093921 005682 0074A 1 D9686. Download the L1 Detected High-Res Dual-Pol (GRD-HD) ProductNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program11

Accessing Sentinel-1 DataNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program12

Opening the Data with the Sentinel Toolbox1. Initiate the Sentinel Toolbox by clicking on its desktop icon2. In the Sentinel Toolbox interface, go to the File menu and select OpenProduct3. Select the folder containing your Sentinel-1 file, and double click on the .zipfile (do not unzip the file; the program will do it for you)National Aeronautics and Space AdministrationApplied Remote Sensing Training Program13

Opening the Data with the Sentinel Toolbox1. The Product Explorer window of theSentinel Toolbox contains your file.Double click on the file to view thedirectories within the file, which containinformation relevant to the image,including:– Metadata: parameters related to orbitand data– Tie Point Grids: interpolation oflatitude/longitude, incidence angle, etc.– Bands: Intensity and amplitude(intensity is the amplitude squared)National Aeronautics and Space AdministrationApplied Remote Sensing Training Program14

Opening the Data with the Sentinel Toolbox4. Worldview image (lower left) shows thefootprint of the image selected.– Note: it is inverted because it is orientedthe same way it was acquiredNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program15

Opening the Data with the Sentinel Toolbox: RGB Image6. Go back to the ProductExplorer tab7. Select the filename of theSentinel-1 dataset8. Select Open RGB ImageWindow to display a colorimage of VV, VH, and VV/VHratioNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program16

Opening the Data with the Sentinel Toolbox: RGB Image9. In the upper, left window, select Pixel Info to see the value and the latitudeand longitude of each pixel in the opened imageNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program17

Preprocessing

Data Preparation: Defining a Subset1. From the top, main menu bar, selectRaster and then Subset accordingto the parameters on the right– From this point on, work only withthe subset image– The created subset is added as anew product on your file windowNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program19

Preprocessing: Geometric and Radiometric Calibration The objective in performing calibration is to create an image where the value ofeach pixel is directly related to the backscatter of the surface This process is essential for analyzing the images in a quantitative way It is also important for comparing images from different sensors, modalities,processors, or images acquired at different timesNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program20

Preprocessing: Radiometric Calibration1. The subset output from the previousstep appears in the ProductExplorer window. Highlight thatoutput by clicking on the file name2. From the top menu, selectRadar Radiometric Calibrateand use the default parameters– this will create a new product withcalibrated values of thebackscatter coefficientNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program21

Preprocessing: Radiometric Calibration The main radiometric distortions are due to:– Signal loss as it propagates– Non-uniform antenna pattern– Difference in gain– Saturation– SpeckleNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program22

Preprocessing: Speckle Reduction Speckle is part of radar images and makes interpretation difficult because the“salt and pepper” effect corrupts information about the surface– There are many techniques to extract radar information from images thathave a lot of speckle– You can use speckle filters, or multilook the image. We will use multilook3. The output from the previous step appears in the Product explorer window(filename ending in Cal). Highlight that output by clicking on the file name4. Select Radar Multilook, then choose the Processing Parameters window.Specify 6 for both number of range and azimuth looksNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program23

Preprocessing: Speckle ReductionCalibrated VVNational Aeronautics and Space AdministrationCalibrated-Multilooked VVApplied Remote Sensing Training Program24

Preprocessing: Geometric Calibration1. The output from the previous stepappears in the Product explorer window(filename ending in Cal ML). Highlightthat output by clicking on the file name2. Select Radar Geometric TerrainCorrection Range-Doppler TerrainCorrection3. In the Processing Parameters tab, usethe default options for the output files,and select UTM/WGS 84 for the mapprojection– This will take 30 minNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program25

Preprocessing: Geometric Calibration Result The main geometric distortions are due to:– Slant Range– Layover– Shadow– Foreshortening The algorithm uses a DEM to make thecorrections The corrected image is in its correctorientationNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program26

Preprocessing: Geometric Calibration Result4. Convert Sigma0 into dB byhighlighting Sigma0 VH and leftclicking5. A menu will pop up. Select Linearto/from dB. Do the same for VV6. Display the dB imagesNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program27

Analysis

Histogram Analysis1. Analyze the imagehistogram in the lower, leftwindow2. Identify the two peaks: thelower one represents waterand the higher onerepresents everything else3. Select the value thatseparates water fromeverything else.– In this case it is -18.85 dBNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program29

Creating a Threshold to Separate Water and Land4. Select Raster Band Math5. Edit the expression so that it reads:– 255*(Sigma0 VHdB -18.85)6. The result will be an image wherewater will have a value of 255. Callthis new image water.National Aeronautics and Space AdministrationApplied Remote Sensing Training Program30

Creating a Threshold to Separate Water and LandNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program31

Example: Processing – Classifying Water and Land1. To change the colors,go to the colormanipulation windowon the bottom left andselect Table2. Assign a color toeach of the threeclassesNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program32

SummaryIt takes different stages in a particular order to generate a product: Data Preparation– Acquire the images– Identify a subsection of the image or create a mosaic, if needed Preprocessing the Image– Radiometric calibration– Filter application to reduce speckle– Geometric Calibration Processing the Image– Generate a map through threshold, supervised, or non-supervisedapproachesNational Aeronautics and Space AdministrationApplied Remote Sensing Training Program33

SAR Processing and Data Analysis Erika Podest . National Aeronautics and Space Administration Applied Remote Sensing Training Program 2 Learning Objective 1. Understand Sentinel Data 2. Perform image preprocessing 3. Analyze SAR imagery to classify land and water . National Aeronautics and Space Administration Applied Remote Sensing Training Program 3 Outline 1. Sentinel-1 Background 2 .