| Desktop SAR Processing: No Longer the Best Kept Secret
With the recent increase in the use of SAR for civilian remote sensing from space, personal SAR processors are providing new opportunities.
By Leo Lightstone Introduction
Desktop software SAR processors are now commercially available. This latest advance in SAR processing technology provides users with the opportunity to fully exploit SAR data while potentially saving thousands of dollars on each purchased scene. These systems are inexpensive, easy to use and run on most standard computer platforms. What used to cost millions of dollars and required a room full of specialized hardware can now be purchased for under 10 thousand dollars and will run on the computer that is probably sitting on the corner of your desk.
The past decade has seen a dramatic increase in the use of synthetic aperture radar (SAR) for civilian remote sensing from space. Presently, there are several fully operational SARs in orbit. An image obtained from any one of these sensors contains hundreds of megabytes of information. However, in the acquisition and formation of an image, the SAR sensor is only one part of a chain (see Figure 1). It is the SAR processor that takes the raw SAR data and "focuses" it to create an image.
Just as image processing underwent a paradigm shift with the computer revolution, so too has the view of SAR processing changed. Once, expensive specially designed hardware was required; now, a full size (100 km x 100 km) SAR image can be processed on a typical Unix workstation in under two hours. The previous need for in-house SAR processing specialists has been eliminated by the SAR processor's user friendly human-machine interface and product support. With present day technology, processing a scene is simply a matter of reading in the data and selecting "go." Advantages and Applications of a DeskTop Software SAR Processor
What are the advantages of a personal desktop SAR processor given that there are ground stations around the world which routinely process SAR images? The answer to this question can be summed up in two words: "flexibility" and "savings."
The major advantage of the software SAR processor is that it provides the user with the flexibility to tune the processing. Customized processing will bring out the image features that are key to a particular application. Another big advantage is the cost benefit that results from purchasing raw rather than processed data. For example, Radarsat International (RSI) quotes raw ERS-1 image data at $925 (CND) per scene whereas a single look processed image costs $2,750. The money saved in purchasing a few scenes alone will cover the entire cost of the SAR processor software. Data re-use also provides a cost saving. Here, a given scene may be processed again and again to meet changing requirements but, only one data set needs to be purchased. The use of one SAR processor for all sensors and all scenes permits consistency in the generated imagery. Finally, the SAR processor provides a research vehicle with which to build new SAR processing tools and develop new applications. History of SAR Processor Development
Unlike raw data from optical sensors, the raw signal collected by the SAR bears little resemblance to an image of the Earth's surface. SAR data looks like the "snowy" picture that can be seen on a television screen when a station is not broadcasting. The SAR processor's role is to manipulate this data into an image.
SAR processing is essentially an application of two dimensional signal processing. Given the number of bits of information in SAR data and the computational expense of focusing an image, the task of creating a SAR processor appears formidable. Two technological achievements conspired to make digital SAR processing a reality. The first of these was the advent of the digital revolution that made fast, accurate digital signal processing (particularly the FFT - Fast Fourier Transform) and digital storage possible. The second key achievement was the recognition that, for standard spaceborne SAR orbits and radar pointing angles, the two-dimensional signal processing required for SAR could be performed as two sequential one-dimensional processing steps, with each step being FFT intensive. The resulting "range-Doppler algorithm" has become the workhorse of SAR processing (although many other approaches are possible). The accuracy, speed and reliability of the digital implementation made these processors suitable for use as part of the ground segment in operational civilian systems. The Image Generation Chain
The process of forming an image starts at the radar which illuminates the Earth's surface with pulses of electromagnetic radiation. These pulses are reflected back to the radar and then transmitted to the ground station, recorded and archived.
The processing of raw data into a SAR image may be performed at the ground station. When an image is required, an order is placed by the user who must choose from a finite number of image types. For each image type, the processing parameters have been pre-selected by the developer of the ground station processor as a compromise to provide reasonable imagery for several applications. This "one size fits all" approach means that the imagery is not optimally processed for most applications. On receipt of the order, an operator at the processing and archiving facility will extract the data from the archive, process the image and ship it to the user.
For the individual, where throughput is not a major constraint and image quality, flexibility and reduced cost are paramount, there is no longer a need to be constrained by the ground station processor. Unprocessed SAR data can be ordered from the archive facility as a "raw data product." With the data and the processor in hand, the user now has the option to process the data with standard settings (just click and go) or to his or her unique specifications. For those who wish to customize their processing, typical image attributes that can be modified include: resolution, focus, speckle reduction factors and a host of other fundamental image parameters.
At first glance it may appear that one can obtain the desired image parameters by processing an existing SAR image with standard image processing tools. Generally this is not true! The SAR processor provides control over the data at the image formation stage thereby maximizing the information retained in the image. As an analogy, think of the SAR sensor as a camera. The SAR processor would then provide the user with control over the focus and other camera settings. A "picture" may be taken over and over again adjusting the settings each time until the desired result is achieved. Example of Some Applications
Many applications benefit from tuning of the SAR processing parameters. For example, custom SAR processing provides significant gains in the areas of interferometry, data fusion, bathymetry and ocean applications (to name just a few). Figures 2, 3 and 4 present some examples of SAR imagery that have been custom processed on the ERGOvista SAR processor developed by Atlantis Scientific. The remainder of this section provides an example of SAR interferometry.
Atlantis Scientific has been using interferometry to study the Kobe area of Japan and the damage resulting from the Hyogoken-nanbu earthquake of January 1995. Figure 3 shows a JERS-1 SAR image of the Kobe region. This image, acquired before the earthquake, and a similar one, acquired after the earthquake, were used together with a digital elevation model to examine the ground displacements caused by the disaster.
The signal data in the Kobe image was corrupted by a ground-based radar operating in the same frequency band as the JERS-1 sensor. This radar interference caused the streaks of noise that cover the image of the land mass. Other problems that arose in some of the data were the appearance of "ghost images" in the scenes and poor matching between the images. (These difficulties are unusual - SAR data is typically easy to work with). Any one of these problems alone made quality interferometric analysis impossible.
Custom SAR processing of the Kobe images was used to address all of the above difficulties. To remove the radar interference a special filtering operation was performed at an intermediate stage in the SAR processing. The "ghost images" were eliminated by correcting an error in the way the SAR data was timed. Finally, the poor matching between the scenes, which was caused by problems in the satellite orbital data, was corrected through custom optimization of the processing parameters. What to Look For Now and In the Future
When considering purchasing SAR processing software there are number of issues to evaluate beyond price. Some of these items are summarized below:
• provide multi-sensor support
• accept data from many ground stations
• provide multi-platform support (PC, Sun, SGI) with image quality independent of platform
• provide increased throughput with standard platform enhancements (i.e. multiple CPUs, increased memory, disk arrays)
• built on a uniform and standard language
• include internal diagnostics, error reporting and on-line context sensitive help
• provide a complete and modern graphical user interface
• provide complete user control over all processing parameters
• provide a full suite of support utilities
• produce radiometrically calibrated imagery
• produce phase preserved, precision imagery
In the future, desk-top SAR processors will continue to have their flexibility enhanced. These new systems will allow the user an even greater range of processing choices through the addition of multiple, selectable, algorithms and advanced software structures.
On the horizon is the use of low cost, special purpose optical computing hardware that may present a major boost to SAR processing. The affordability and practicality of such components compared with the continually evolving digital computers remains to be seen.
Regardless of their form, it is clear that desk-top software SAR processors have entered a new era as commercial products. These systems will enable users to build and perfect existing applications while developing new ones. Personal SAR processors will continue to play a key role in revealing the wealth of information contained in SAR data while making the use of this data more affordable. Suggested References
An excellent description of SAR and SAR processing for those with an engineering background is provided by Charles Elachi in his book titled, "Spaceborne Radar Remote Sensing: Applications and Techniques," IEEE Press, 1988. This book presents a very good discussion of all of the issues without being too mathematical.
For those who are interested in understanding SAR processing at the detailed level, the book by J. Curlander and R. McDonough, "Synthetic Aperture Radar: Systems and Signal Processing," John Wiley & Sons, Inc., 1991, is highly recommended. Acknowledgement The artist's rendition of Radarsat (used in Figure 1) supplied courtesy of Canadian Space Agency. About the Author:
Leonard Lightstone is the manager of the SAR Processor Product Development group at Atlantis Scientific, Ottawa, Canada. He has worked for the past 11 years in the area of radar and space-based systems. He may be reached at 613-727-1087 or e-mail at [email protected]. Back |
