Experiences with the South African VHF Synthetic Aperture Radar

1
Experiences with the South African VHF Synthetic
Aperture Radar
R.T. LordA.J. WilkinsonM.R. Inggs
2
Contents

• South African SAR (SASAR) System Overview
• VHF SAR Characteristics
• Semi Desert and Forest Measurements
• Radio Frequency Interference Suppression
• Repeat-Pass Interferometry
• Results
• Discussion of Results
• Conclusions
• Future Work

• R.T. Lord, A.J. Wilkinson and M.R. Inggs

3
SASAR System Overview

• Centre frequency 141 MHz, Wavelength 2.13 m
• Bandwidth 12 MHz (resolution 12 m)
• Transmit power 1 kW
• Fully polarimetric
• Antenna 4 elements, inclined monopoles
• Flight platform DC3 (C47 Dakota)
• Wide range of look angles

Typical SASAR imaging geometry - to scale.
R.T. Lord, A.J. Wilkinson and M.R. Inggs
4
Antenna Structure on the DC3
Antenna
R.T. Lord, A.J. Wilkinson and M.R. Inggs
5
Typical SASAR System Parameters
R.T. Lord, A.J. Wilkinson and M.R. Inggs
6
VHF SAR Characteristics

• Good foliage canopy penetration gt bald earth
  topography mapping and the detection of vehicles
  under bush cover
• Reasonable ground penetration capabilities
  expected in arid regions gt geological and
  hydrological mapping applications
• Forest stem volume measurements
• Very sensitive to large-scale man-made structures
  such as buildings, fences and power lines
• Ship detection for maritime patrol applications
• Comparison of the imagery from a VHF radar and a
  higher frequency radar provides additional
  information

R.T. Lord, A.J. Wilkinson and M.R. Inggs
7
Semi Desert and Forest Measurements

• The SASAR system has flown over two different
  terrain types
• 1. a flat semi-desert area near the town of
  Upington, and
• 2. over a mountainous, forested coastal area
  near the town of Hermanus
• Flat, unvegetated desert terrain provides almost
  no backscattered signal for larger incidence
  angles
• Coastal scene gt large returns from heavily
  vegetated areas over the entire range extent were
  obtained, with many features of interest
• Only the steepest mountain slopes provide
  significant backscatter gt forest biomass
  measurements in this type of terrain look
  promising
• Interferometric mapping of ground layer

R.T. Lord, A.J. Wilkinson and M.R. Inggs
8
RFI Measured with the SASAR System

• RFI measured with the SASAR system over Hermanus
  near Cape Town
• The graph shows the magnitude averaged range
  spectrum of 1000 range lines for the H-receive
  polarisation data, with a PRF of 136 Hz
• Frequency extent is 24 MHz
• Note that this graph shows that the antenna and
  receive filters have not adequately limited the
  receive bandwidth to 12 MHz

R.T. Lord, A.J. Wilkinson and M.R. Inggs
9
SASAR Image Contaminated with RFI
VHF-band image of the vicinity of Upington, South
Africa, degraded by RFI. The flight path is along
the horizontal axis, with near range towards the
bottom of the image.
R.T. Lord, A.J. Wilkinson and M.R. Inggs
10
SASAR Image Cleaned with LMS Adaptive Filter
R.T. Lord, A.J. Wilkinson and M.R. Inggs
11
Interferometric Experiment Objectives

• Explore the potential of repeat-pass
  interferometry at VHF frequencies
• Investigate suitable baseline geometries
• Understand the limitations
• Compare VHF band to other bands (e.g. ERS C-band)
• Investigate possible applications
• coherence mapping
• change detection
• height mapping
• etc.

R.T. Lord, A.J. Wilkinson and M.R. Inggs
12
Interferometric Simulator Output

• Noise-free interferogram, noisy interferogram and
  coherence
• 100m vertical baseline
• Observing a horizontal surface

R.T. Lord, A.J. Wilkinson and M.R. Inggs
13
Baseline Selection

• Conclusion Vertical baselines of between 100m
  and 300m seemed suitable
• Selection aided by design curves and simulations

Possible baselines
H13800
100 m
H23700
200 m
H33500
300 m
R.T. Lord, A.J. Wilkinson and M.R. Inggs
14
Data Acquisition and SAR Processing

• Data location - Hermanus/Botrivier coastline,
  South Africa
• Very mountainous, forested region - significant
  backscatter
• Three passes
• Time separation of about 25 minutes
• Vertical baselines of 105 m, 206 m and 311 m
• Swath width about 28 km
• Processed to SLC images using in-house
  range/Doppler processor
• RFI suppression using notch filter
• Interferometric processing performed with GAMMA
  software

R.T. Lord, A.J. Wilkinson and M.R. Inggs
15
Image Registration

• Tie point offsets should increase linearly in
  flight direction
• However, azimuth offset displays step-like
  nature
• indicates missing range lines
• Bilinear warping function is not sufficient for
  registration
• results in bands of high correlation
• Development of registration algorithm with
  arbitrary amount of warping freedom
• performs localized warping on smaller patches
• Significant improvement in registration and hence
  coherence

R.T. Lord, A.J. Wilkinson and M.R. Inggs
16
Aerial Photo of Hermanus/Botrivier Area
R.T. Lord, A.J. Wilkinson and M.R. Inggs
17
SASAR Image of Hermanus/Botrivier Area
R.T. Lord, A.J. Wilkinson and M.R. Inggs
18
Map and Intensity Image of Hermanus/Botrivier Area
Highlighted contour spacing 100m
R.T. Lord, A.J. Wilkinson and M.R. Inggs
19
Simulated and Real Coherence Image
R.T. Lord, A.J. Wilkinson and M.R. Inggs
20
Simulated and Real Interferogram (unflattened)
Simulated
Real
R.T. Lord, A.J. Wilkinson and M.R. Inggs
21
Simulated and Real Interferogram (Flattened)
R.T. Lord, A.J. Wilkinson and M.R. Inggs
22
Discussion of Results

• Drop-outs occurring in data recording system
• Bunching-up or stretching-out of range lines due
  to platforms deviation from nominal flight path
• Motion compensation performed to mid-range
• Position data possibly contains errors in the
  order of a few wavelengths
• RFI suppression - decrease in SNR
• Non-parallel flight paths - discrepancy with
  simulated interferogram

R.T. Lord, A.J. Wilkinson and M.R. Inggs
23
Conclusions

• High coherence in bright regions only (i.e. SNR
  limited)
• Strong signals obtained in densely vegetated
  areas (scatterers comparable to wavelength)
• Limited potential for height mapping
• Low coherence patches in areas containing no
  vegetation
• Phase unwrapping complicated by patch-like
  structure
• Maybe good in heavily vegetated regions
• Less sensitive to temporal decorrelation than
  microwave bands
• Fringe information does, however, complement a
  backscatter image
• Especially consider that at VHF, measured
  backscatter is very insensitive to local slope
  variations. Not much information regarding local
  slope can be inferred from a single VHF SAR image
• RFI removal necessary

R.T. Lord, A.J. Wilkinson and M.R. Inggs
24
Future Work

• Fix problem with azimuth sampling
• We suspect problem with data recorder, i.e.
  missing lines
• Could also be related to motion compensation
• Slave PRF to ground speed
• Time-stamp received data
• Attempt phase unwrapping of interferogram
• DEM reconstruction and evaluation
• Studies of coherence properties of scenes

R.T. Lord, A.J. Wilkinson and M.R. Inggs