Professional Documents
Culture Documents
COPYRIGHT NOTICE
Software copyrighted by: PCI Geomatics Enterprises Inc. 50 West Wilmot Street Richmond Hill, Ontario L4B 1M5 CANADA Telephone number: (905) 764-0614 The Licensed Software contain material that is protected by international Copyright Law and trade secret law, and by international treaty provisions, as well as by the laws of the country in which this software is used All rights not granted to Licensee herein are reserved to Licensor. Licensee may not remove any proprietary notice of Licensor from any copy of the Licensed Software.
RESTRICTED RIGHTS
Canadian Government Use, duplication, or disclosure is subject to restrictions as set forth in Public Works and Government Services Canada Supplemental General Conditions Licensed Software 4003. U.S. Government Use, duplication, or disclosure by the Government is subject to restrictions set forth in subparagraph (b)(3) of the Rights in Technical Data and Computer Software clause of DFARS 252.227-7013 or subparagraph (c)(1) and (2) of the Commercial Computer SoftwareRestricted Rights clause at 48 CFR 52.227-19 as amended, or any successor regulations thereto. The Government agrees: (a) if the Licensed Software is supplied to the Department of Defense (DoD), the Licensed Software is classified as Commercial Computer Software and the Government is acquiring only RESTRICTED RIGHTS in the Software as that term is defined in Clause 252.227-7013(c)(1)(ii) of the DFARS (48 CFR 52.227-19); and (b) if the Licensed Software is supplied to any unit or agency of the United States Government other than DoD, the Governments rights in the Licensed Software will be as defined in Clause 52.227-19 (c)(2) of Supplement to the FAR. PCI, PCI Geomatics, PCI and design (logo), Geomatica, Committed to GeoIntelligence Solutions, GeoGateway, FLY!, OrthoEngine, RADARSOFT, EASI/PACE, ImageWorks, GCPWorks, PCI Author, PCI Visual Modeler, and SPANS are registered trademarks of PCI Geomatics Enterprises, Inc. Sentinel is a registered trademark of SafeNet, Inc. Copyright 1990-2007 Info-ZIP. All rights reserved. InstallAnywhere is a registered trademark of Macrovision Corporation. ESRI, ArcGIS, ArcGIS Server, ArcGIS Server Image Extension, and ArcMap are trademarks, registered trademarks, or service marks of ESRI in the United States, the European Community, or certain other jurisdictions. All other trademarks and registered trademarks are the property of their respective owners.
Table of Contents
SAR processing with Geomatica............................................................................................... 1 Radar Processing Workflows .................................................................................................. 2
Module 1: SAR processing in OrthoEngine .......................................................................... 1-3 Starting OrthoEngine ........................................................................................................ 1-4 Lesson 1.1: Creating a Project and Importing Radar Data .................................................. 1-5 Creating a Project ............................................................................................................. 1-6 Adding images to the project ............................................................................................ 1-9 Saving the Project ........................................................................................................... 1-10 Lesson 1.2: DEM Extraction .............................................................................................. 1-11 Lesson 1.3: Generating Orthorectified Images .................................................................. 1-17
Module 2: Calibrating your data ............................................................................................. 2-1 Lesson 2.1: Extracting a calibrated backscatter image........................................................ 2-2 Starting Geomatica ........................................................................................................... 2-3
Module 3: Applying radiometric enhancements ................................................................... 3-1 Lesson 3.1: Applying SAR speckle filters ............................................................................ 3-2 Lesson 3.2: Computing SAR texture measures ................................................................... 3-7
Module 4: Classifying SAR data ............................................................................................. 4-1 Lesson 4.1: Segmenting SAR data...................................................................................... 4-2 Lesson 4.2: Classifying SAR data........................................................................................ 4-8
Module 5: Working with the SAR Polarimetry Workstation ................................................. 5-1 SAR Polarimetry Workstation components ....................................................................... 5-2 Lesson 5.1: Creating a PCIDSK file from a Polsar dataset.................................................. 5-5 Exporting the Polsar dataset to PCIDSK........................................................................... 5-5 Lesson 5.2: Applying SAR Filters in Focus .......................................................................... 5-7 Lesson 5.3: Using the SAR Polarimetry Target Analysis tool .............................................. 5-9 Starting the SPTA ............................................................................................................. 5-9 Selecting targets and creating response plots ................................................................ 5-10
PCI Geomatics
Table of Contents
Module 6: Classifying polarimetric data in Modeler ............................................................. 6-1 Lesson 6.1: Ingesting and filtering polarimetric data............................................................ 6-2 Lesson 6.2: Calculating polarimetric discriminators ............................................................. 6-6 Lesson 6.3: Unsupervised classification of POLSAR data ................................................ 6-11
ii
PCI Geomatics
Introduction
Welcome to the SAR Processing with Geomatica training course. This course is designed for experienced users of geospatial software. This course will introduce you to the radar analysis tools available in PCI Geomatica software. This manual contains six modules. Each module contains lessons that are built on tasks that you are likely to perform in your analysis of radar imagery. They provide instruction for using the software to carry out essential processes while sampling key Geomatica applications and features.
Each module in this book contains a series of hands-on lessons that let you work with the software and a set of sample data. Lessons have brief introductions followed by tasks and procedures in numbered steps. The lessons provide you an opportunity to work with several Geomatica applications and to carry out different tasks using OrthoEngine and Focus. Your overall goal is to become familiar with the tools available for Synthetic Aperture Radar (SAR) analysis, and to see how you can use Geomatica in your own work. The scope of this guide is confined to the core PCI software applications included in the Geomatica suite. However, some radar remote sensing concepts are reviewed in the modules and lessons. The six modules in this course provide instructions for a range of activities associated with processing SAR data. This course includes the following modules: Module 1: SAR processing in OrthoEngine Module 2: Calibrating your data Module 3: Applying radiometric enhancements Module 4: Classifying SAR data Module 5: Working with the SAR Polarimetry Workstation Module 6: Classifying polarimetric data in Modeler
The data you will use in this course can be found in the Radar Data folder supplied on the CD accompanying your software. It is recommended that you copy this data to your hard disk.
PCI Geomatics
At the orthorectification step, you have different resampling options. Your choice will depend on where you want to proceed with your data. If you know which filter type and size will produce the best results for your image, you can apply the filter at the same time as you orthorectify the image. Otherwise, you can use Nearest Neighbor resampling during orthorectification, then filter and manipulate the orthorectified image to obtain the desired results. The dataset used in this course has varied relief and will be used to demonstrate the workflow for active terrain. Tools for working with flat terrain will also be examined.
PCI Geomatics
Module
Creating a Project and Importing Radar Data DEM Extraction Generating Orthorectified Images
The Satellite Orbital math model is a rigorous model developed by Dr. Toutin at the Canada Centre for Remote Sensing to compensate for distortions such as sensor geometry, satellite orbit and attitude variations, and earth shape, rotation, and relief. This model is used for high-resolution and low-resolution imagery from a variety of optical sensors and radar images such as ASAR, RADARSAT, ERS-1 and JERS1. The computed math model calculates the position and orientation of the sensor at the time when the image was taken. The Radar Specific model uses the additional parameters in the orbit data to diminish the amount of ground control points (GCPs) required. The extra parameters maintain the positional accuracy and high levels of detail in the model, but reduces the number of GCPs needed to few or none. This math model does not use tie points because each scene is computed using the GCPs of that scene only. If you have more than eight well-distributed GCPs, both Toutin's model and the Radar Specific model will provide similar results.
PCI Geomatics
1-3
Starting OrthoEngine
Windows systems To start OrthoEngine on Windows systems:
1. Click the Start menu and select: All Programs > PCI Geomatics > Geomatica Vx.x > OrthoEngine Alternatively, if the Geomatica Toolbar is running, click the OrthoEngine button. The OrthoEngine window opens.
Unix systems
Figure 1-1:
OrthoEngine window
1-4
PCI Geomatics
Lesson 1.1
The objective of this lesson is to create a project and import a pair of RADARSAT images. To create this project, you require: Radar images in their raw data format Map projection information
The Radar Specific model is designed for stereo DEM extraction and image orthorectification of ASAR and RADARSAT data. Compared to the Rigorous Satellite model for radar data, the Radar Specific model uses the additional parameters in the orbit data to diminish amount of ground control points (GCPs) required. The extra parameters maintain the positional accuracy and high levels of detail in the model, but the number of GCPs required is reduced to few or none. In addition, the model uses satellite ancillary data provided in the CEOS data structure to calculate the satellite position and extract the slant range distance to target. The addition of ground control points, although not necessary, will refine the model and improve its accuracy. Unlike the Rigorous Satellite model, the Radar Specific model does not use tie points (TPs) because each scene is computed using the GCPs of that scene only. If you have more than eight well-distributed GCPs, both Toutin's model and the Radar Specific model will provide similar results.
PCI Geomatics
1-5
Creating a Project
OrthoEngine works on a project-by-project basis. You must open an existing project or create a new project before you gain access to the functions within OrthoEngine. In this lesson, you will create a new Radar project using RADARSAT data.
9. Click OK. The Project Information window closes and the Set Projection window appears.
Figure 1-3:
Set Projection window
1-6
PCI Geomatics
Output projection
GCP projection
PCI Geomatics
1-7
Tip
To modify the projection information, reopen the Set Projection window.
Warning
Changes to the projection mid-project will render any existing orthophotos invalid.
1-8
PCI Geomatics
Caution
If you save satellite data from the CD onto a hard disk before reading it to a PCIDSK file, it is important that you maintain the naming structure of the folders as they appeared on the CD. If the structure or folder names are changed, you may encounter errors.
Now that you have created the project file, you will read the raw data to a PCIDSK file. The project file will then contain the file name and location of each input image.
2. On the Data Input toolbar, click Read CD-ROM data. The Read CD-ROM window opens. 3. For the CD Format, select RADAR (CEOS). 4. Beside CD image file name, click Select. A File Selection window opens. 5. Locate the Radar Data folder and open the SCENE01-STD2A folder. 6. Select the file DAT_01.001 and click Open. The path and file name are updated in the window. 7. For the Requested Channels, click 1. If you have Single Look Complex data, you would click both 1 and 2 to output the real and imaginary layers to two separate channels. 8. For the SAR Type, select RADARSAT. 9. For the PCIDSK File name, type std2.pix. A report file will also be created. 10. Click Read. When the file is read, it is part of your project. 11. Repeat steps 4 to 10 to read data from the SCENE01-STD6A folder and name the file std6.pix. You have now added a second radar image to your project.
PCI Geomatics
1-9
Tip
If you need to revert to the backup file, rename the backup file so that it uses the .prj extension. OrthoEngine can load this project file in the normal way.
1-10
PCI Geomatics
Lesson 1.2
DEM Extraction
This lesson describes how to convert two RADARSAT-1 scenes to their epipolar projections. The epipolar pair will be used for extracting and geocoding a DEM. To complete this lesson you require: The two RADARSAT-1 scenes, std2.pix and std6.pix The project file radar.prj
DEM extraction
The process of generating a digital elevation model (DEM) consists of several steps: Converting the raw images into epipolar pairs. Epipolar images are stereo pairs that are reprojected so that the left and right images have a common orientation, and matching features between the images appear along a common x axis. Extracting DEMs from the overlap between the epipolar pairs. The resulting DEMs are called epipolar DEMs. They are not georeferenced at this stage. Geocoding the epipolar DEMs and stitch them together to form one DEM.
PCI Geomatics
1-11
2. On the DEM From Stereo toolbar, click Create Epipolar Image. The Generate Epipolar Image window opens.
Figure 1-6:
Generate Epipolar Images window
Now you will select the scenes to generate the epipolar images.
1-12
PCI Geomatics
The Epipolar Pairs table lists one epipolar pair: std2 and std6. 4. In the Options section, specify the Working Cache. This is the amount of RAM that you want to allocate to create the epipolar pairs.
Note
It is recommended that you NOT exceed 50% of the available RAM. Specifying more than half the RAM available on your system may significantly reduce performance.
5. Select a Downsample Factor of 2. This lets you downsample the epipolar images so that two adjoining pixels and lines will form one pixel in the epipolar image. Adjust this value if you see noisy features in your image that you do not want to see in your DEM. 6. For the Downsample Filter, select Average. 7. Click Save Setup. This saves the options chosen for batch processing with Automatic DEM Extraction. 8. Click Close. Now that you have set up the epipolar image generation, you will set up for DEM extraction.
Tip
CCRS provides an interactive Web tool to help you select a RADARSAT stereo pair. See http://ccrs.nrcan.gc.ca/resource/tutor/advisor/index_e.php.
OrthoEngine uses image correlation to extract matching pixels in the two images, then uses the sensor geometry from the computed math model to calculate x, y, and z positions. The ASAR/RADARSAT Specific model is based on orbital information instead of ground control points. Orbit information always references an ellipsoid, and the ellipsoid number is taken from the projection information defined in the file. A DEM extracted from satellite imagery using this math model is based on an ellipsoidal model of the earth, not on the geoid. The difference between elevations relative to the ellipsoid and those relative to the geoid may be significant. PCI Geomatics 1-13
You will begin by extracting the DEM from the epipolar pair to create an epipolar DEM, then you will geocode the DEM based on the math model.
The epipolar pairs created in the last lesson are listed in the Stereo Pairs selection table. The initial status of the pair is Offline because you set up the options for generating the epipolar pair, but did not generate them. They will be generated and the DEM will be extracted all in one step.
1-14
PCI Geomatics
This controls the size of the pixel in the final DEM relative to the input images. The higher the number, the larger the DEM pixel will be, and the faster the DEM is processed. 8. Leave the Use Clip Region checkbox cleared. 9. Select Fill Holes & Filter. This will automatically interpolate the data over failed areas and filter elevation values based on adjacent pixels. 10. Leave the Delete Epipolar Pairs After Use checkbox cleared. 11. Click Extract DEM. 12. When the DEM extraction is complete, click Close. Now that you have extracted the epipolar DEM, you will geocode the DEM based on the math model.
2. To locate the Input DEM File, click Browse. 3. From the Radar Data folder, select dem_std2_std6.pix and click Open. The DEM channel is automatically selected as channel 3 and the default Failure and Background values are displayed. 4. Accept the default Input Window setting of Full Image. 5. To specify the Output DEM File, click Browse and locate the Radar Data folder. 6. For the File name, type geocoded_dem.pix and click Open. The path and file name appear in the Output DEM File field. 7. For Pixel Spacing, enter 50m. This specifies the pixel spacing of the epipolar DEM. 8. For the File holes option, select Yes.
PCI Geomatics
1-15
9. Click Geocode DEM. 10. When the process is complete, click Close.
Tip
The Convert DEM Datum tool under the Utilities menu allows you to convert your DEM from Ellipsoid to MSL or MSL to Ellipsoid.
1-16
PCI Geomatics
Lesson 1.3
This lesson describes how to set up and perform orthorectification. For this lesson you require: The radar.prj file for the two radar scenes The orangedem.pix digital elevation model
Data correction
Orthorectification is the process of using a rigorous math model and a digital elevation model (DEM) to correct distortions in raw images, as shown in Figure1-10. The rigorous math models, such as the Aerial Photography or Satellite Orbital math models, provide a method for calculating the position and orientation of the sensor at the time when the image was taken. The DEM is a raster of terrain elevations. The quality of the orthorectified image is directly related to the quality of the rigorous math model and the DEM. A poorly computed math model, an inaccurate DEM, or a DEM incorrectly georeferenced to the math model will cause errors in the orthorectified images.
Figure 1-10:
Using sensor geometry and a DEM to orthorectify imagery
When working with the Radar Specific model, ground control points (GCPs) are optional. Without GCPs, the model is calculated according to the satellites positioning information. The addition of ground control points, although not necessary, will refine the model and improve its accuracy. This means, however,
PCI Geomatics
1-17
that as long as you have a DEM, or have extracted a DEM from stereo radar images, you can still orthorectify your radar data without any ground control.
2. On the Ortho Generation toolbar, click Schedule ortho generation. The Ortho Image Production window opens.
Figure 1-12:
Ortho Image Production window
3. Under Available Images, use the SHIFT key to select both images and click the arrow button to move the images under Images to process. The images are processed in the order that they are listed. 4. Under Images to Process, select std2. By default, the ortho image will be named ostd2.pix. You may type a different file name in the Ortho Image section File field.
1-18
PCI Geomatics
2. From the Radar Data folder, select orangedem.pix and click Open. The DEM File Window opens. 3. For the Background elevation, enter -3. This represents the No Data pixels in the DEM. For DEMs generated by OrthoEngine, the background elevation defaults to -150. Other DEMs may have different background elevation values that you must know before they can be used. If you do not know the background value, click DEM Info in the DEM File window. The window displays the three lowest and three highest values in the DEM. 4. For Datum, select Mean Sea Level. The DEM will be automatically converted from mean sea level to ellipsoid. 5. Click OK.
Sampling Interval
Resampling
PCI Geomatics
1-19
The Land Cover Correction is an experimental algorithm designed to compensate for the scattering of the radar signal due to the texture and composition of the surface, particularly for areas with uniform vegetation. The amount of Land Cover Correction is controlled by a coefficient (exponent) ranging between 0 and 2.0. Coefficient values between 0.5 and 1.0 produce the best results for most images. Since the results depend on the sensor, the beam mode, and the image, finding the best result for your project will involve some experimentation.
Filter Size
The Filter Size parameter controls the size of the kernel by specifying the number of pixels (width) in the X box and the number of lines (length) in the Y box. Different filter sizes greatly affect the quality of the processed images. If the filter is too small, the noise filtering algorithm is not effective. If the filter is too large, subtle details of the image are lost in the filtering process. The minimum size for the kernel is 3x3 pixels. A 7x7 kernel usually gives the best results. Note that the Filter Size parameter becomes enabled when a filter is applied during resampling. The radar images are supplied in one of two Image Formats: Power or Amplitude. Power is the sum of the squares of the real and imaginary values of the complex pixel values in the radar image. Amplitude is the square root of Power. Most radar images are supplied in the Amplitude format to preserve the values. You identify the format used with your images in the Image Format list. Note that the Image Format is only valid when a Radar Specific filter is applied during resampling. You use the Number of Looks to estimate noise variance and to control the amount of smoothing applied to the image. In theory, the correct value for the Number of Looks should be the effective number of looks of the input radar image, or close to the actual number. Using a smaller value for the Number of Looks leads to more smoothing, and a larger value preserves more image features. In this example, you will perform the orthorectification and also apply the Gamma filter. The resulting image could then be used for intensity analysis.
Image Format
Number of Looks
Tip
Filtering should be performed during the correction step to avoid multiple resampling of the data, which can degrade and reduce interpretability of the imagery.
1-20
PCI Geomatics
6. For the Image Format, select Amp. 7. For the No. of Looks, select 4. The Image Format and the Number of Looks of the radar image are usually recorded on the CD jacket or in the format specifications provided by the data vendor.
Figure 1-13:
Ortho Image Production window with parameters
Now that you have entered all the necessary parameters, you can generate the ortho images.
2. Click Close. The message Ortho done appears beside each image in the Available Images section, indicating that the original images are now orthorectified. PCI Geomatics 1-21
Tip
This is a good time to save your project file.
In generating the orthos, you selected the Radiometric Correction option. The output orthorectified images are in units of sigma nought (radiometrically calibrated), have the radiometric terrain correction applied, and are now stored in 32-bit channels. Now that you have corrected your radar images and applied a SAR filter to remove speckle, your data can be used for the statistical analysis or modeling of SAR reflectivity. In the modules that follow, you will examine additional tools in Geomatica for extracting calibrated backscatter values and applying radiometric enhancements.
1-22
PCI Geomatics
PCI Geomatics
1-23
1-24
PCI Geomatics
Module
Radiometric calibration
Calibration is the process whereby the digital values of a received remotely sensed image can be related to physical quantities of the scene, such as brightness (reflectivity), phase and location. The goal of radiometric calibration is to account for all of the contributions in the radiometric values which are not due to target characteristics, so that the backscatter values of targets can be compared to one another or a reference. Most data products are radiometrically calibrated by the data supplier prior to being distributed. The last step in processing is the application of look-up tables (LUTs) during calibration from floating point data to digital numbers (DN) to ensure the best use of the 8 or 16 bits on storage media. These look-up tables apply a fixed offset and a range-dependent gain function to the data prior to generating the final image output. The scaling used can vary for each scene, making it difficult to directly relate information between scenes (that is, radiometric and temporal signatures). The look-up tables are provided with the data and are used by Geomatica to extract the calibrated image. Proper SAR processing should first convert pixel values (DNs) to sigma-0, then use that for any further analysis. Geomatica has tools for converting from DNs to both beta-0 and sigma-0.
PCI Geomatics
2-1
Lesson 2.1
Radar brightness
The SARBETA algorithm generates a radar brightness channel from both RADARSAT and ASAR data. This brightness channel allows comparison within and between radar scenes by inverting scaling and offsets applied during processing. This reverts the data to its calibrated form and outputs the data as brightness values (beta nought). Radar brightness is a measure of the total energy returned from an area on the ground projected into a given image pixel (similar to radiance in optical imaging). The use of brightness tends to be somewhat limited to purists who may want to look at the data in this form possibly to verify sensor calibration and performance.
Radar backscatter
The SARSIGM algorithm is used to revert commercially processed SAR data to its calibrated radar backscatter component. It uses the orbit, offset, and scaling segment created during data input, and, for RADARSAT data, the incidence angle segment created by SARINCD, to produce a 32-bit real image which is radiometrically calibrated and which more accurately represents the original signal amplitudes (sigma nought). SARSIGMA calculates sigma nought, which is the measure of the strength of radar signals backscattered towards the radar antenna from a distributed scatterer, usually expressed in dB. It is a measure of the SAR reflectivity of terrain (similar to reflectance in optical imaging). It is a dimensionless number, giving the mean backscatter from an area of one square meter on the surface of the Earth. Sigma nought is dependent on the properties of the scatterer (ground cover) and usually varies significantly with incident angle, radar, or wavelength, and polarization. SARSIGM assumes flat terrain and should be applied to images with flat terrain before further processing. For RADARSAT images, SARINCD must be run before SARSIGM to create the incident angle array segment. For supported ASAR images it is automatically created on import.
Incident angles
SARINCD produces a table of incident angles that correspond to the table of gain scaling values produced when an image is read using CDSAR. The incident angle is the angle between the line of sight from the radar to an element of an imaged scene, and a vertical direction characteristic of the scene.
2-2
PCI Geomatics
Incident angle is the most important parameter describing the relative geometry between the radar and the observed scene.
Note
SARINCD is not required for ENVISAT ASAR images. CDASAR extracts incident angle information from the ESA-format image and stores it in the Incidence Angle segment.
Up to this point, you have been using the SAR-specific tools available in OrthoEngine. In this lesson you will access SAR processing tools available through the Algorithm Librarian in Focus.
Starting Geomatica
Windows systems To start Geomatica on Windows systems:
1. From the Start menu, select: All Programs > PCI Geomatics > Geomatica Vx.x > Geomatica The Geomatica Toolbar and a Focus window open.
Unix systems
Figure 2-1:
Geomatica Toolbar
PCI Geomatics
2-3
Figure 2-2:
Focus window
2-4
PCI Geomatics
Figure 2-3:
Algorithm Librarian
2. Expand the Radar Analysis folder. The tree expands to display a series of subfolders. 3. In the Radar Analysis category, expand the SAR Radar Analysis folder. A list of SAR processing algorithms is displayed. 4. Double-click the SARINCD algorithm. The SARINCD Module control panel opens.
Figure 2-4:
SARINCD Module control panel
PCI Geomatics
2-5
2-6
PCI Geomatics
Figure 2-5:
SARSIGM Module control panel
2. In the Input Ports section of the Files tab, click the + beside Input Layer(s). 3. From std6.pix, select channel 1: CDSAR BSQ Band 1. 4. Click the + beside SAR Offset Layer. The tree expands to show the array segments in std6.pix. 5. Select array segment 3: SAR Offset. 6. Click the + beside SAR Gain Layer and select array segment 4: SAR Gain. 7. Click the + beside SAR Incident Angle Layer and select array segment 6: SAR Incident Angle. 8. Click the + beside Satellite Orbital Ephemeris Layer and select segment 2: CDSAR Orbital Ephemeris Data. 9. Click the + beside GCP Layer and select segment 5: GCPs from the Orbit segment. Now that you have specified the Input Ports, you will specify the Output Ports. The Viewer-Grayscale option is selected by default.
PCI Geomatics
2-7
Note
If a path for the output file is not specified, the file is saved in the user folder.
6. Click the + beside Satellite Orbital Ephemeris Layer. 7. Select the Untitled2.pix check box and copy and paste the path and file name specified above. The ephemeris data will be written to a new layer within the std6_backscatter.pix file. 8. Repeat steps 6 and 7 for the GCP Layer. The GCPs will be written to a new layer within the std6_backscatter.pix file.
Image format
There are three formats in which the radar values can be stored: amplitude, decibels, and power. Amplitude values (where amplitude equals the square root of the radar backscatter coefficient) have a positive range of real values. Decibel (dB) values are a logarithmic scaling of power (sigma nought). Decibels produce a range of real numbers that may include negative and positive values. Power values (where power equals the radar backscatter) have a positive range of values and are the square of amplitude.
2-8
PCI Geomatics
Tip
See Appendix A for information on converting the format of radar values.
PCI Geomatics
2-9
2-10
PCI Geomatics
Module
Radiometric enhancements
The goal of radiometric enhancements is to improve the interpretation of the radiometric information in an image using speckle and spatial filters. These filters can reduce speckle, detect edges, analyze texture, and visually enhance the image. Image variance, or speckle, is a granular noise that is inherent in SAR imagery. Speckle gives a grainy, salt and pepper appearance and is the dominating factor in radar imagery. Speckle filters are primarily used on radar data to remove high-frequency noise (speckle), while preserving high-frequency features (edges). Texture is one of the important characteristics used to identify objects or regions of interest in an image. Unlike spectral features, which describe the average tonal variation in the various bands of an image, textural features contain information about the spatial distribution of tonal variations within a band.
PCI Geomatics
3-1
Lesson 3.1
Coherent signal scattering in SAR data often causes image speckle or salt-andpepper effects. Speckle is inherent in most SAR images, and can inhibit accurate image interpretation. There are several types of speckle filters. They fall into two general categories: non-adaptive and adaptive. With non-adaptive filters, the parameters of the whole image are considered. They do not take into account the local properties of terrain backscatter or the nature of the sensor. Examples of non-adaptive filters are the Mean, Median, Edge Detection, and Sieve filters. Adaptive filters accommodate changes in local properties of the terrain backscatter. They modify the image based on statistics extracted from the local environment of each pixel. The main advantage of adaptive filters is that they maintain an accurate estimate of the backscattering coefficient inside homogeneous (stationary) areas while preserving edge and texture structure in nonstationary scenes. Examples of adaptive filters are the Frost, Lee, Gamma and Touzi filters. Adaptive filters are designed specifically for SAR images. The ideal speckle-reduction filter will reduce the speckle but have minimal loss of information. The filter should preserve radiometric information in homogeneous areas and preserve radiometric information and spatial signal variability (textural information) in textured areas. FSPEC (SAR Speckle Filters) is an algorithm that lets you select which SAR speckle filter you wish to apply to your data. It allows you to select from many SAR speckle filters in a single panel. The filter algorithms require that the signal represents power. If the input image is in amplitude format, each gray level will be squared to derive power and, finally, the square root will be applied to the filtered result. In this lesson, you will apply and compare several adaptive SAR speckle filters.
3-2
PCI Geomatics
With the Frost filter, the unspeckled pixel value is estimated for each pixel by computing the weighted sum of the center pixel value, the mean value, and the variance calculated in a circular kernel surrounding the pixel. The enhanced Frost filter minimizes the loss of radiometric and textural information.
PCI Geomatics
3-3
Figure 3-1:
FSPEC Module control panel for Enhanced Frost filter
Note
You must enter correct parameters for the image files you are working with. The number of looks and the image type information is available in the format definition included with your data.
8. Click Run. The Enhanced Frost filter runs and the results are shown in the Focus viewer.
3-4
PCI Geomatics
Gamma filter
The Gamma filter assumes that the (unspeckled) intensity of the underlying scene is gamma-distributed. The filter minimizes the loss of texture information better than the Frost and Lee filters within gamma-distributed scenes. The Gamma filter is suitable for a wide range of gamma-distributed scenes, such as forested areas, agricultural areas, and oceans.
7. For Image Type, select POW. 8. For Filter Type, select Gamma. 9. Select a Filter X size and a Filter Y size of 7. The filter kernel size will be 7x7 pixels. 10. For the Number of Looks, enter 4. 11. Click Run. The Gamma filter runs and the results are displayed in the Focus view area.
PCI Geomatics
3-5
Touzi filter
Unlike other filters, the Touzi filter adapts the size and shape of the filter processing window to signal nonstationarity. It also has a multi-resolution ratio edge detector for better filtering of contours and edges. This is effective when applied to fine structures such as roads and trails which are generally smoothed out by other filters.
3-6
PCI Geomatics
Lesson 3.2
SARTEX
SARTEX creates a set of texture images from a single channel in an input image. The radar-specific measures are based on pixel value statistics in a window surrounding each pixel. The extracted texture measures may be used as input features to classification algorithms. SAR-specific texture measures offered by this module account for radar image formation and statistical properties of radar speckle. Therefore it is essential that no prior filtering has been performed on the input channel. Texture measures extracted by SARTEX are based on SAR-specific statistics of pixel values in the rectangular window of user-specified dimensions. They complement measures extracted in the module TEX, based on a gray level cooccurrence matrix (GLCM), and in the module HISTEX, based on histogram of pixel values in a window. The three types of measures can be combined for further analysis, such as segmentation and classification of the image. The best window size depends on image characteristics (terrain roughness, land cover type, imaging parameters). For large windows, some blocky artifacts may appear in texture images if a small, bright target is fully enclosed by the window. Therefore, small or moderate window sizes are recommended in most cases. For gentle terrain, texture analysis should be performed on the raw image without calibration to brightness. For moderate/rough terrain, it should be performed on radiometric terrain corrected orthoimages. In the latter case, it could be calculated in the raw image domain, but the texture would be contaminated by terrain effects.
PCI Geomatics
3-7
Figure 3-3:
SARTEX Module control panel
3. In the Input Ports section of the Files tab, click the + beside Input: Image Layer. 4. From the std6_backscatter.pix file, select Calibrated Radar Backscatter. 5. In the Output Ports section, right-click Untitled.pix file and click Browse. A File To Save window opens. 6. Locate the Radar Data folder. 7. For the File name, type sartex.pix. 8. Click Save. The path and file name are updated in the SARTEX Module control panel.
3-8
PCI Geomatics
Figure 3-4:
SARTEX parameters
In Table 3.1, each SAR texture measure is listed with a brief explanation. Table 3.1: SAR-Specific Texture Measures Texture Measure VI Explanation This is a ratio of the mean of squared intensity to the mean intensity squared. It is applicable when pixel intensity has a Gaussian distribution. This is a ratio of mean intensity to the squared mean amplitude. It is applicable when pixel amplitude has a Gaussian distribution. This is a difference of the mean value of the squared intensity logarithm and the square of the mean intensity logarithm. It is applicable when pixel intensity has a lognormal distribution. This is a normalized log measure of texture. It approximates K distribution when the number of looks in the SAR image is large.
VA
VL
PCI Geomatics
3-9
2. For the SAR Texture Measurement Parameter, select VI. 3. For the Filter X Size, select 5. 4. For the Filter Y Size, select 5. 5. For Image Format, select POW. It is important to select the correct image format for the input layer. Any required conversion are performed internally to use the correct values for each computed texture measure. 6. Click Run. A Progress Monitor opens on your desktop. When the algorithm has finished executing, the results are displayed in the Focus view area and you can close the TEX Module control panel. If you like, try calculating another SAR texture measure and compare it to the VI measure.
Tip
To solve a specific problem (distinguishing between several land cover types for example), the usual approach is to compute all texture measures and use the measure that provides the best result.
3-10
PCI Geomatics
Module
4
Segmenting SAR data Classifying SAR data
The classification of SAR data using standard classification algorithms developed for the multispectral classification of optical data does not usually yield successful results because of the heavy speckle affecting SAR imagery. The SARSEG and SARCLASS algorithms are used together to segment the SAR image into regions of similar mean backscatter values. Region-based classification is then performed by the SARCLASS algorithm, based on a set of training regions for each class.
PCI Geomatics
4-1
Lesson 4.1
SARSEG
SARSEG segments images into regions and is meant to be used in conjunction with SARCLASS to classify SAR images. The input to SARSEG is a 32R image channel. The input channel should contain SAR backscatter values in the power or intensity scaling. For some images, the amplitude scaling may also work well. No pixel value conversions are applied in SARSEG. The resolution of the input image should correspond to the physical resolution of the sensor on the ground, as opposed to the pixel size of the standard (oversampled) SAR image products. The output of SARSEG is in two channels in one file. The 16U channel labels each pixel using the ID of the region containing the pixel, and the 32R channel sets each pixel to the value of its region mean. The output channels are intended to be used as inputs for the module SARCLASS. SARCLASS assigns individual regions to one of several classes, based on a set of training regions for each class. Because the resolution of the input image should correspond to the physical resolution of the sensor rather than to the pixel size, you will begin by downsampling the std6_backscatter.pix file.
4-2
PCI Geomatics
Figure 4-1:
RESAMP Module control panel
Note
If a path for the output file is not specified, the file is saved in the user folder.
PCI Geomatics
4-3
2. For both the Output Pixel X and Y Size, enter 25. This is the actual resolution of the sensor. 3. For the Resample Method, select Average. The input pixel values are averaged to create the new layer at lower resolution. 4. Click Run. The RESAMP algorithm runs and the results are displayed in the Focus viewer. Close the RESAMP Module control panel.
4-4
PCI Geomatics
One of the output channels from SARSEG is a 16-bit unsigned channel that labels each pixel by the ID of the region to which it belongs. This means there is a maximum of 65,535 regions, or segments, that can be created. For this reason, you will work with a subset of the downsampled backscatter image.
PCI Geomatics
4-5
2. For the Maximum Number of Iterations, enter 10. By default, the algorithm executes until the total standard deviation of the image stops decreasing. By setting this parameter, the algorithm will stop after the number of specified iterations 3. For the Edge Detection Threshold, use the default value of 4.6.
4-6
PCI Geomatics
Increasing this value will decrease the number of accepted edges, therefore decreasing the total number of regions found in the image. 4. For the Sobel Filter Edge Selection Threshold, use the default value of 2.0. This specifies the threshold used to select edge pixels after the Sobel filter stage. Increasing this value will decrease the number of accepted edges, therefore decreasing the total number of regions found in the image. 5. For Do Small Region Detection, select YES. This will create many regions of only one or two pixels, likely producing slightly better results. However, due to limitations of a 16U channel, a maximum of 65535 regions can be created. Not performing Small Region Detection will dramatically reduce the number of segments created. 6. For the Region Merging Sieving Threshold, use the default value of 0. This is the size in pixels of the smallest region not to be merged into a neighbor. This parameter will decrease the total number of regions, but will result in the loss of some small but potentially distinct targets. By setting this value to 1 or lower, region merging will not be performed. 7. Click Run. The SARSEG algorithm runs and the results are displayed in the Focus viewer.
PCI Geomatics
4-7
Lesson 4.2
SARCLASS
SARCLASS classifies homogeneous image regions defined for a SAR image, based on the image values within each segment. A supervised classification is performed, with the required classes provided as parameters to SARCLASS. Each region is assigned to a class or left unclassified, depending on the input parameters. All pixels of all regions assigned to a given class are marked as "on" bits in the output bitmap corresponding to this class. The input file for SARCLASS is an output file from SARSEG or other segmentation algorithm. It must contain an image layer and a layer with unique region IDs. If SAR classification is requested, the image layer is assumed to contain positive SAR intensity or radar backscatter values, and no pixel value conversion is performed. For the purpose of SARCLASS, the suitable estimates of radar cross section are: pixel intensity (power); calibrated radar brightness beta-0; radar backscatter sigma-0; and radar backscatter gamma-0. The classification in SARCLASS is performed on a per-region basis, as opposed to the standard per-pixel classification. This means that classification measures dependent on pixel value histograms within regions (median, mode) cannot be used. Moreover, such parameters are not applicable to SAR images. Two classification measures supported in SARCLASS are a SAR-specific measure, and a region mean. Before you perform the classification of the segmented radar image, you must collect training information.
4-8
PCI Geomatics
PCI Geomatics
4-9
3. In the Input Ports section of the Files tab, click the + beside Input Image Channel. 4. From the std6_resamp_sub.pix file, select the Subset from std6_resamp layer. 5. For the Input Region Numbers Channel, select the sarseg: region IDs layer. 6. For the Training Point Layer, select the New Point Layer. Now that you have specified the input, you will configure the Output Port. One output bitmap will be created for each class that is defined. The Viewer option is selected by default.
4-10
PCI Geomatics
2. For the Field Name, enter Class_Name. 3. For the Classification Measure, select SAR. The SAR specific measure is a Maximum Likelihood classifier, which calculates the probability of a region belonging to a given class. With the SAR measure, all training regions must have their mean value greater than zero. For the MEAN classification measure, a region is assigned to the class which has its mean closest to the region's mean. 4. For the Classification Threshold, enter 1.0. This specifies the classification measure threshold that each region must meet to be assigned to a class. It also implicitly controls the creation of the "zero class" of unclassified pixels. All input regions are classified, unless a threshold is specified. Practical values range from 0.5 to 1.5, but will depend on the image and the classification measure. 5. For the Report Mode, select SARCLASS.RPT. 6. For the Classification Report Type, select LONG. This provides detailed statistics of all training and classified regions, grouped by class. 7. Click Run. The SARCLASS algorithm runs and the results are displayed in the Focus viewer.
PCI Geomatics
4-11
4-12
PCI Geomatics
Module
Creating a PCIDSK file from a Polsar dataset Applying SAR Filters in Focus Using the SAR Polarimetry Target Analysis tool
Polarimetry fundamentals
The study of polarimetry refers to the science behind and the techniques relating to the analysis of the polarimetric properties of electromagnetic waves, and the scatterers of these waves. When four polarization combinations are used in orthogonal pairs, the complete scattering properties of radar scatter can be determined for a specific frequency, incidence angle, and azimuth direction. The extraction of information such as terrain classification is an important component of polarimetry. In single polarized images, the return is proportional to the target crosssection. The amount of return that the radar system gets is dependent on the target, the orientation of the target as well as the radar system, and the polarization. Most natural targets scatter the radar signal and give a large return in HV, whereas most artificial targets only give large returns in HH and VV, since a majority of these targets are bright in nature.
Polarimetry applications
There are numerous advantages of using polarimetric data for image analysis. Agricultural applications include crop type identification, crop condition monitoring, and soil moisture measurement. Forestry applications include biomass estimation and species identification. Other applications of polarimetric data include geology, hydrography, oceanography and costal zone mapping.
PCI Geomatics
5-1
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
Figure 5-1:
The Cloude-Pottier CBDT window
5-2
PCI Geomatics
Cloude and Pottier (1997) grouped the different scattering mechanisms in 17 different classes, indicated by the Cloude-Pottier definition table. Classes are chosen based on a particular scattering mechanism and are not specific to a particular dataset.
All Cloude-Pottier classes can be grouped into eight scattering mechanisms. The first eight classes have anisotrophy values ranging from 0 to 0.5, indicating azimuthally symmetrical surfaces. The last eight classes have anisotrophy values ranging from 0.5 to 1.0, indicating features with a high probability of a dominant (one-dimensional) scattering pattern. Classes 1 & 9: Double-bounce scattering in a high-entropy scattering Classes 2 & 10: High entropy vegetation scattering (for example, forest canopy)
PCI Geomatics
5-3
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
Classes 3 & 11: Medium entropy multiple scattering Classes 4 & 12: Medium entropy vegetation scattering Classes 5 & 13: Medium entropy surface scattering Classes 6 & 14: Low entropy multiple scattering Classes 7 & 15: Low entropy dipole scattering Classes 8 & 16: Low entropy dipole scattering It is important to keep in mind that although there is some degree of arbitrariness in how the boundaries between these classes are defined, they are closely linked with physical scattering mechanisms, independent of the training dataset.
SPW can be used to analyze the following types of polarimetric SAR data products: ASAR single-look complex (APS); Precision (APP); Ellipsoid geocoded (APG); CV-580 and X-band SAR single-look complex (SLC) and multi-look complex (MLC) AIRSAR multi-look complex (MLC) RADARSAT-2
A number of other products are supported in SPW; this is a list of the most commonly used data products. SPW can also analyze any polarimetric SAR dataset that can be imported into a PCIDSK file with polarimetric SAR-specific metadata (use PSINGEST to import a compatible dataset into a PCIDSK file).
5-4
PCI Geomatics
Lesson 5.1
The objective of this lesson is to create a PCIDSK file for use in the SPTA tool. To create this project, you require: the product.xml file from the Flevoland folder and supporting files
4. Select Calibration Type: Sigma Nought and click OK. The Raster Display Option window opens, presenting the options of displaying the raster with georeferencing from the file, or from a math model. 5. Select File, and click OK. A message window appears, asking if you wish to display the raster North Up or Raster Up. 6. Select North Up, and click OK. The file is displayed in Focus. 7. From the Files tree, right-click the product.xml file to display the context menu, and select Translate (Export) File. The Translate (Export) File window opens. 8. In the Destination file field, click Browse, then navigate to the Flevoland data folder. 9. In the File name field, type flevoland, then click Save. 10. In the Translate (Export) window, select the PIX:PCIDSK output format. 11. For the Source Layers, click the Select All button.
PCI Geomatics
5-5
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
All of the layers of product.xml are selected. 12. Click >Add>. The selected layers from the product.xml file are transferred to the Destination Layers section of the Translate (Export) File window.
Figure 5-4: The Translate (Export) File window.
13. Click Translate. A progress bar appears, indicating that the file is being exported to PCIDSK format. 14. When the progress bar disappears, click Close to close the Translate (Export) File window.
5-6
PCI Geomatics
Lesson 5.2
This lesson describes how to apply a specific SAR filter to the imagery.To complete this lesson you require: The file flevoland_tiled256.pix
PSBOXCAR
The PSBOXCAR algorithm applies a boxcar filter to a polarimetric SAR dataset. This filter is used to increase the effective number of looks in a single-look or multi-look dataset. This step is important for processing because the PSBOXCAR module will reduce the noise inherent in radar images, while preserving the polarimetric information. This process makes features in the imagery easier to interpret. The resulting file will be imported into the SPTA tool.
PCI Geomatics
5-7
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
4. In the File name field, type psboxcar.pix and click Save. 5. In the Input Params1 tab, type 7 for both the Window size in X direction and Window size in Y direction. 6. Leave the Decimate Output checkbox cleared. 7. Click Run. A progress bar appears indicating that the PSBOXCAR algorithm is running on the specified image. 8. When the module runs to completion, use File > Open to open psboxcar.pix. Compare psboxcar.pix to flevoland_tiled256.pix. 9. From the Focus File menu, select Exit. A warning dialog appears, asking if you would like to save the changes to the unnamed project. 10. Click No. The psboxcar.pix file has been created. This file, along with the flevoland_tiled256.pix file, is now ready to be used in the SPTA.
5-8
PCI Geomatics
Lesson 5.3
In order to complete this module, you require: The flevoland.pix file created in Lesson 5.1 The psboxcar.pix file created in Lesson 5.2
PCI Geomatics
5-9
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
2.
5-10
PCI Geomatics
5. Select North Up, and click OK. The flevoland.pix file is displayed at 1:1 resolution in the Target Selection window with channels 9, 10 and 12 mapped to RGB, respectively.
Figure 5-7:
The flevoland.pix image loaded in the Target Selection window (1:1 resolution)
6. In the Target Selection window, click Zoom to Overview. 7. In the SPTA window, click the + beside Graphical Output. The Graphical Output options are displayed. 8. For the Target Selection Mode, select the Arbitrary Region option. 9. From the Shape drop-down list, select Rectangle. 10. In the Target Selection window, use your mouse to digitize a small polygon over the harbor area in the image.
PCI Geomatics
5-11
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
Figure 5-8:
Harbor region in the flevoland.pix image
11. In the SPTA window, from the Graphical Output options, select the Response Plot tab. 12. From the Polarization Response Plot Input parameters, select co-polarized for type, and normalized for the scaling. 13. Click Plot. The co-polarized response plot is created and opens in a new window. The values are high in the center of the plot, indicating a strong response in HHHH. This type of response is an indication of horizontal features. We also see moderately high values at the edges of the plot. Given the area that was used to create this plot, we could infer that these values are indicating the presence of ships with masts.
5-12
PCI Geomatics
Examine the co-polarized plot for the harbor area and close the plot. Evaluate their responses at 0 degrees (horizontal polarization), 90 (vertical polarization). The maximum response for this feature is received when ellipticity and orientation angles are both 0, which means linear (ellipticity=0) horizontal (orientation=0) polarization. The peak in this graph means that the harbour has a dominant scattering pattern at 0 degrees, which makes this target easier to discriminate from other features.
Figure 5-10:
Co-polarized plot for harbor region
14. In the SPTA window, click the + beside Numerical Output. The numerical output is displayed. Examine the values for the arbitrary region target (rectangle), specifically the average target pixel intensity values for HH, HV, VH and VV.
PCI Geomatics
5-13
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
Figure 5-11:
Approximate zoom area
Figure 5-12:
Approximate location of the communication tower feature
5-14
PCI Geomatics
5. From the Target Analysis window, in the Target Selection Mode, select the Square Neighbourhood option. In the Target Selection window, the pointer turns to a crosshair with a square attached to it. 6. Select the control tower feature by clicking on the feature. 7. In the SPTA window, from the Graphical Output options, select the Response Plot tab. 8. From the Polarization Response Plot Input parameters, select co-polarized for type, and normalized for the scaling. 9. Click Plot.
Figure 5-14:
Co-polarized response plot for high elevation feature
10. In the SPTA window, click the + beside Numerical Output. The numerical output is displayed. Examine the average target pixel intensity values for HH, HV, VH and VV.
PCI Geomatics
5-15
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
2. Navigate to felvoland.pix, and click Open. The Raster Display Option window opens, presenting the options of displaying the raster with georeferencing from the file, or from a math model. 3. Select File, and click OK. A message window appears, asking if you wish to display the raster North Up or Raster Up. 4. Select North Up, and click OK. The felovland.pix file is displayed at 1:1 resolution in the Target Selection window, with channels 9, 10 and 12 mapped to RGB, respectively. 5. Map the RGB channels to 1, 2 and 3, respectively. 6. Apply a Linear enhancement to the image. The image is updated with 1, 2, 3 for RGB respectively, and the Linear enhancement is applied. 7. In the Target Selection window, click Zoom to Overview. 8. In the SPTA window, from the Target Selection Mode, select the Arbitrary Region option. 9. From the Shape drop-down list, select Polygon. 10. In the Target Selection window, use your mouse to digitize a small polygon over a forest in the image.
Figure 5-15: Example
of a forest sample area in flevoland.pix
5-16
PCI Geomatics
11. In the SPTA window, select the + beside Graphical Output to expand the Graphical Output options. 12. Select the Response Plot tab. Under the Polarization Response Plot parameters, choose co-polarized for type, and normalized for the scaling. 13. Click Plot.
Figure 5-16: Response plot for random features
The co-polarized response plot is created and opens in a new window. The values are high in both the vertical and horizontal planes, indicating random features. Most natural features will result in a plot of this type. Examine the co-polarized plot for the harbor area and close the plot.
PCI Geomatics
5-17
SAR Processing with Geomatica - Module 5: Working with the SAR Polarimetry Workstation
5-18
PCI Geomatics
Module
Ingesting and filtering polarimetric data Calculating polarimetric discriminators Unsupervised classification of POLSAR data
Classification algorithms are grouped into supervised and unsupervised methods, although some algorithms combine features from each category. For supervised classification, an image analyst identifies land cover classes in a scene. This process is known as training the data. The training data can be chosen from the scene itself, or from scenes that possess similar characteristics. Once the training is complete, supervised classification algorithms automatically assign each pixel to a class based on the predetermined class means or boundaries. For unsupervised classification, the unsupervised classification algorithm has no prior information about the scene content or the landcover types present in the imagery. An unsupervised classification algorithm will examine the parameter space for each scene and will automatically define class boundaries based on homogeneous pixels. When the classification algorithm has run to completion, the image analyst must identify each class manually. Classification of polarimetric data follows the same logic, where unsupervised classification algorithms (such as PSCLOPOT and PSUSWIS) use polarimetric signatures derived by arbitrarily partitioning the feature space of the data, whereas supervised classification algorithms (such as PSSWIS) use training classes provided as bitmaps.
PCI Geomatics
6-1
Lesson 6.1
PSINGEST
PSINGEST imports the specified polarimetric SAR dataset into the PCISDK format to increase performance during subsequent processing. Both the image data and metadata are imported. The entire image or a subset of the image may be imported. The imported image is stored in one of the POLSAR matrix formats, which is automatically selected according to the data contained in the image.
PSBOXCAR
6-2
PCI Geomatics
Figure 6-1:
Modeler environment with PSINGEST, PSBOXCAR, and Export modules
Tip
To quickly find algorithms without expanding categories in the Algorithm Librarian, click Find and type in the name of the algorithm that you wish to find. Click Find Next in the Find Algorithm panel to go to the next algorithm with the matching name.
PCI Geomatics
6-3
This optional parameter creates a subset of the data that is 400 pixels by 800 lines. The upper-left corner of the subset window will be 1000 pixels from the left of the original image, and 5000 pixels from top. 4. For Data Calibration ensure that the default, Sigma Nought, is selected. 5. Accept the default Database Layout of band, and Pyramid Options of near. 6. Click Accept.
Figure 6-3:
The PSINGEST module control panel
Now that you have specified the input, you will configure the Export module. 6-4 PCI Geomatics
Note
If a path for the output file is not specified, the file is saved in the user folder.
PCI Geomatics
6-5
Lesson 6.2
PSEABA
The PSEABA algorithm creates an image of four parameters for a POLSAR dataset. The four parameters are entropy, alpha angle, beta angle, and anisotrophy; these parameters characterize the properties of partially coherent scattering.
PSPOLDIS
The PSPOLDIS algorithm calculates a number of polarimetric discriminators for a POLSAR dataset. These discriminators are based on the polarimetric synthesis. This algorithm reconstructs the orientation and ellipticity angles that a completely polarized wave would have for the different features in the image.
6-6
PCI Geomatics
Figure 6-6: Modeler canvas with the PSEABA, PSPOLDIS, Export, and Viewer modules added
3. On the PSBOXCAR module, click the Output Filtered Scene port, then click the Input Polarimetric SAR Scene port on the PSEABA module. A green pipe joins the two modules. 4. On the PSBOXCAR module, click the Output Filtered Scene port, then click the Polarimetric SAR Scene port on the PSPOLDIS module. A green pipe joins the two modules. 5. On the PSEABA module, click the Output Cloude-Pottier Results port, then click the Port on the EXPORT module. A green pipe joins the two modules. 6. On the PSPOLDIS module, click the Output Polarimetric Discriminators port, then click the Raster Port on the Viewer module. A green pipe joins the two modules.
PCI Geomatics
6-7
6-8
PCI Geomatics
2. From the Modeler toolbar, click Run. The model runs through each of the modules. When the model runs to completion, the message Execution Complete appears at the bottom of the Modeler window. 3. When the model has run to completion, open the flevoland_pseaba_mod.pix image in Focus to view the results from the PSEABA module. The results from the PSPOLDIS module appear in a Modeler Viewer window. Figure 6-10 shows the output files from PSEABA (in Focus), and PSPOLDIS (in the Viewer window).
Figure 6-10:
The output images created by the PSEABA and PSPOLDIS modules.
In the flevoland_pseaba_mod.pix output file, the values of channels 1, 2, and 3 are mapped to Entropy, Anisotropy, and Alpha angle, respectively. This means that the range of red pixels indicate pixels with low to high entropy, and so on.
PCI Geomatics
6-9
The output file contains the derived polarimetric discriminators and related angles at every pixel.
6-10
PCI Geomatics
Lesson 6.3
Note
For this module, you will require the model from the previous lessons
PCI Geomatics
6-11
Figure 6-12:
Unsupervised classification modules
7. On the PSBOXCAR module, click the Output Filtered Scene port, then click the Input Polarimetric SAR Scene port on the PSVANZYL module. A green pipe joins the two modules. 8. On the PSVANZYL module, click the Output Classified Image port, then click the port on the first EXPORT module. A green pipe joins the two modules. 9. On the PSBOXCAR module, click the Output Filtered Scene port, then click the Input Polarimetric Scene port on the PSCLOPOT module. A green pipe joins the two modules. 10. On the PSCLOPOT module, click the Output Cloude-Pottier Classified Image port, then click the port on the second EXPORT module. A green pipe joins the two modules. 11. On the PSBOXCAR module, click the Output Filtered Scene port, then click the Input Polarimetric SAR Scene port on the PSFREDUR module. 12. On the PSFREDUR module, click the Output Classified Image port, then click the port on the third EXPORT module. 13. On the PSBOXCAR module, click the Output Filtered Scene port, then click the Input Polarimetric SAR Scene port on the PSUSWIS module. 14. On the PSFREDUR module, click the Output Classified Image port, then click the Input Freeman-Durden or Cloude-Pottier Classified Image port on the PSUSWIS module. 15. On the PSUSWIS module, click the Output Wishart Classified Image port and click the port on the fourth EXPORT module. Ensure that your model appears similar to the model in Figure 6-13.
6-12
PCI Geomatics
Figure 6-13:
SPW workflow
PCI Geomatics
6-13
6-14
PCI Geomatics
PCI Geomatics
6-15
Figure 6-15:
Viewing output from the PSVANZYL module
Original psvanzyl.pix image created from the PSVANZYL unsupervised classification method.
The psvanzyl.pix image after PCT editing, showing all classes generated from the PSVANZYL unsupervised classification method.
6-16
PCI Geomatics
14. To view the number of classes that the unsupervised classification module computed, right-click the psvanzyl.pix file in the Maps tree and select Histograms. The Histogram Display panel is displayed. The x-axis shows the number of classes that exist in this classification. In this case, there are 4 classes.
Figure 6-16:
The Histogram Display panel, showing the number of classes generated from the unsupervised classification module, PSVANZYL
15. Repeat steps 2 to 15 to view the classification results generated from PSCLOPOT and PSUSWIS.
Figure 6-17:
Classification generated from the PSCLOPOT module (16 classes)
PCI Geomatics
6-17
Figure 6-18:
Classification generated from the PSUSWIS module (9 classes)
6-18
PCI Geomatics
PCI Geomatics
6-19
6-20
PCI Geomatics
Appendix
The conversion to other scalings can be easily performed in EASI Modeling. The models provided here assume %1 is the original input channel and %2 is the new output channel. All involved channels must be 32R to accommodate pixel values smaller than 1.0 or even negative (in decibels). The image channel must exist in the database .pix file prior to running the model.
4. From the Input File list, select the input file. 5. Click in the model box and enter one of the models below. To convert from Amplitude to Power: %2 = %1^2 To convert from Amplitude to Decibels: if( %1 > 0 ) %2 = 10 * LOG10( %1^2 ) To convert from Power to Amplitude: %2 = %1^0.5 To convert from Power to Decibels: if( %1 > 0 ) %2 = 10 * LOG10( %1 ) To convert from Decibels to Power: %2 = EXP10( %1 / 10.0 )
PCI Geomatics
A-21
SAR Processing with Geomatica - Appendix A: EASI Modeling to convert radar value formats
To convert from Decibels to Amplitude: %2 = (EXP10( %1 / 10.0 ))^0.5 6. To view the results in the Focus window, select the Display Result(s) checkbox. 7. Click Run.
A-22
PCI Geomatics