Introduction to a SAR System
One way of collecting imagery from space on cloudy areas or at darkness conditions is to fly a Synthetic Aperture Radar (SAR) working at microwave frequencies. As the satellite moves along its orbit, the SAR looks out sideways from the direction of travel, acquiring the radar echoes (See SAR geometry) which return from the strip of the Earth's surface under observation.
This strip, called the radar swath, runs parallel to, and to one side of, the satellite's ground track. ESA's first generation of Earth Observation satellites, ERS-1 and ERS-2 launched in 1991 and 1994 respectively, fly at a mean altitude of 785 km, and have a look angle of 23 degrees from local vertical, and a 100 km swath width. In contrast to images taken by classical visible and infrared 'camera like' sensors, images collected by a SAR are severely unfocused when in their raw form and are unusable without additional data processing. In fact, considerable processing is required to generate a focused image with high spatial resolution (See the Development of a SAR processor section). In the case of ERS-1, resolutions in the order of 5 metres along track (parallel to the flight direction) and 8 meters in slant-range (perpendicular to the flight direction) are achieved. After slant to ground range projection, the ERS-1 ground range resolution is approximately 20 meters. Improved imaging capabilities are provided by the Advanced SAR (ASAR) flown on the Envisat-1 Polar Platform. ASAR has an electronically steerable antenna for acquiring image data with selectable look angles and swath positions. In particular, the so called "wide swath" new mode of ASAR can collect image data over a 400 km. swath by periodically changing the look angle so that the radar beam is switched between five adjoining sub-swaths. This sort of scanning is called ScanSAR, and has the advantage of providing a much larger coverage at the expense of degrading the spatial and radiometric resolution of the final image (See images). ScanSAR techniques are also used in the American SIR-C and the Canadian RADARSAT missions. For all the missions and modes described above, the echoes are digitised On-Board and immediately sent to ground for processing. We provide here below the key parameters of the popular ERS-1 satellite. For a better comprehension, compare this to the geometry figure of a SAR system.
Typical ERS-1 Parameters are shown here below:
| Carrier frequency (fo): | = 5.3 GHz => = wavelength = 5.66 cm. |
|---|---|
| Sampling frequency (fs): | = 18.96 MHz. |
| Pulse chirp modulation: | = rect [T] exp (2PI [fo t + (Bw/2T) t2]) |
| Bandwidth (Bw) | = 15.5 MHz. |
| Pulse duration (T) | = 37.12 msec. => 704 samples at 18.96 Mhz) |
| 1 / PRF (PRF=1678 Hz) | ~ 600 msec. |
| Echo window: | ~ 300 msec. => 5616 samples (~100 Km swath) |
| Echo delay (2R / c) | ~ 5600 msec. =>(Rx 9 pulses after Tx) |
| Satellite Altitude | ~ 780 Km (equator) ~ 810 Km (South pole) |
| Slant Range distance (R) | ~ 830 Km (near) ~ 870 Km (far). |
| Range distance from nadir: | ~ 250 Km (near) ~ 350 Km (far)=> Swath 100 Km |
| Incidence angle | ~ 20 deg. near range - 26 deg. far range. |
| Data rate(5 bits I,5 bits Q) | ~ 5616 x 1678 x 10 ~ 100 Mbit / sec. |
| Relative speed target-sensor (vst) | ~approximately 7000 meter / sec. |
|---|---|
| Distance inter-pulse | ~ 4 m. => ~27000 echoes for 108 Km. image |
| Azimuth illumination time (3 dB)~5 Km. | =>~ 1200 echoes |
Antenna dimensions: Lrange = 10 meter , Lazimuth = 1 meter
| Raw data set: | ~27000 x 5616 x 2 (I,Q) => ~ 300 MByte |
|---|---|
| SLC Image (100Km-100Km) | =25000 x 5000 x 4 (I,Q) => ~ 600 MByte |
| FDC Image (100Km-100Km) | = 6300 x 5000 x 2 (detected,3 looks) = 63 MByte |
| PRI (100km-100km) | = 8000 x 8000 x 2 (detected,3 looks) =128 MByte |
| Quicklook (100Km-100Km | = 1000 x 1000 x 1 (detected,>30looks) = 1 MByte |
Note that in the case of ERS-1 and Envisat-ASAR the raw data rate is about 100 Mbit/sec. This is directly related to the range and azimuth pixel spacing, which depend on system constraints.
Development of a SAR processor
The on-board SAR system receives the backscattering echoes from the ground, and then these echoes are transformed to baseband and digitised. This digitised data is called RAW DATA. The raw data is then stored temporarily in Solid State Mass Memories or immediately sent to Earth via the telemetry downlink at some 100 Mbit/s for on-ground processing.
The sort of star that appears in the RAW DATA of the figure above corresponds to a very bright (artificial) point target that looks enhanced with respect to all other points. Actually, all points have the same form or impulse response. The raw data is then processed (compressed) in range (horizontal) direction, and each point target takes a sort of very fine C shaped long line. This phenomenon is called range migration, and needs to be compensated in the processor. Data is then processed in azimuth or fligth (vertical) direction. This process is usually performed for small blocks or tiles due to the limitation of memory resources in computers. The generation of big images thus requires to build the images as a mosaic.
everal high-resolution algorithms exist to perform the range and azimuth compression. They mainly differ in the way they perform the range migration compensation. Those algorithms are:
- Range-Doppler
- W-K
- Chirp Scaling Algorithm
In order to perform a sub-pixel accuracy quality assessment, very bright (artificial) point target are interpolated, and then the profiles of their 3-D impulse response analysed.
Among others, additional processing stages, necessary to produce images with square pixels that can be matched to GIS systems, are:
- multi-look summation combined with azimuth resampling
- slant to range projection combined with range resampling.
- geo-referencing:
- removal of altitude influence.
- rotation of the image to correct the orbit inclination
The SAR instrument acquires samples at a constant increment of time. This gives a constant slant range pixel spacing, however, due to the variation in the incidence angles, an interpolation is needed to compensate the variation that appears on the ground-projection.