SASAR VHF Imagery 

The signal processing for the system is performed using the G2 SAR Processor, developed over a number of years at the UCT RRSG. This fully motion-compensated, airborne SAR processor is based on the range-Doppler algorithm with the main modules written in C and using Python as the glue language.

Our analysis and simulation has shown that the range-Doppler approach provides good quality imagery for the modest resolution requirements of the current SASAR VHF system. In addition, we have developed an efficient extension to range-Doppler processing which allows for higher resolution processing at VHF, although this is not required for operational use of the current system.  Other approaches suitable to high resolution, low frequency processing include the range migration approach and the time domain approaches traditionally requiring more computation, but allowing for a more natural wide-beam motion compensation implementation.

When processing multiple images with the G2 processor, efficient use is made of our parallel cluster, Gollach, using the Mosix extension of the Linux kernel. In addition, efforts are underway to produce a parallel version of G2 using PVM (Parallel Virtual Machine) which will speed up the processing of individual SASAR images.
 

Hermanus, South Africa 1999-11-29 (Latest!)

These runs were along the Southern coast of South Africa in the Hermanus area near Cape Town.


Upington, South Africa 1999-09-02

All images with motion compensation using IMU and DGPS, unless specified. These images display good focus. The town of Upington may be seen as the bright region to the left of the image. The regions away from the town have very low radar backscatter at SASAR frequencies and correspond to flat, unvegetated, desert regions. The bright, horizontal feature near the base of the image is the airport fence. Other fence structures are also visible as bright linear features.

Upington, South Africa 1999-09-04 (Interferometry)

A number of runs at different heights were performed. This allowed for some repeat-pass interferometric processing to be performed. Note that these results are very preliminary. Improvements to the image registration algorithm should improve the results. In addition, the Upington scene is not well suited to intereferometry due to the extremely low backscatter from the flat, unvegetated desert surrounds. The town areas and fences provide the regions of highest coherence. The processing was performed on VV data. Note that the radar was not performing as well on this day as for the 1999-09-02 images. Note also that no correction has been made in these images for the tracks not being accurately parallel.

Tzaneen, South Africa 1999-07-21

This is a forested and mountainous area. Returns are seen right to the far swath (24 km). During these flights a number of problems with the radar were encountered, but some interesting imagery was still obtained. These are the first flights where the DGPS and pulse expander were used.

Flight from Pretoria to Cape Town 1999-01-15 (historical interest only)

Inertial data available in two dimensions (no height change data). No DGPS available and monochromatic pulses used. Smooth conditions reported on flight. Click here for a map of the flight path for the data recorded during the flight.
 

On-line images (preliminary - no fine tuning performed)

In all images, the pixels are approximately square, slant range is displayed with near range at bottom. The flight direction (azimuth) is from right to left (as if looking out of the window for the right-looking antenna). Unless indicated, no motion compensation has been applied.

In many images, the bright horizontal line near the bottom is the nadir return and faint repeated horizontal lines are artifacts of the interference suppression caused by processing of the nadir return. Interference shows up as bright vertical lines. Short, bright horizontal lines in some images are thought to be man-made linear structures whose orientation happens to be parallel to the flight path (some ground truth work to follow...).

Flight North of Pretoria 1999-01-13 (historical interest only)

First radiating flight, 1999-01-13, area north of Cullinan mine, very rough conditions, no inertial data available. No DGPS available and monochromatic pulses used. The on-line images have the following parameters:

On-line images (note very preliminary - no fine tuning of the processing applied):



Enquiries / suggestions about this site or the SASAR system and signal processing.

Up to SASAR home page

This page last updated by tb  2005-07-07  (C) UCT 1999-2000