Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.advisorGray, Douglas-
dc.contributor.advisorPreiss, Mark-
dc.contributor.authorPincus, Paul-
dc.description.abstractCoherent change detection (CCD) is an established technique for remotely monitoring landscapes with minimal vegetation or buildings. By evaluating the local complex correlation between a pair of synthetic aperture radar (SAR) images acquired on repeat passes of an airborne or spaceborne imaging radar system, a map of the scene coherence is obtained. Subtle disturbances of the ground are detected as areas of low coherence in the surface clutter. This thesis investigates extending CCD to monitor the ground in a forest. It is formulated as a multichannel dual-layer coherence estimation problem, where the coherence of scattering from the ground is estimated after suppressing interference from the canopy by vertically beamforming multiple image channels acquired at slightly different grazing angles on each pass. This 3D SAR beamforming must preserve the phase of the ground response. The choice of operating wavelength is considered in terms of the trade-off between foliage penetration and change sensitivity. A framework for comparing the performance of different radar designs and beamforming algorithms, as well as assessing the sensitivity to error, is built around the random-volume-over-ground (RVOG) model of forest scattering. If the ground and volume scattering contributions in the received echo are of similar strength, it is shown that an L-band array of just three channels can provide enough volume attenuation to permit reasonable estimation of the ground coherence. The proposed method is demonstrated using an RVOG clutter simulation and a modified version of the physics-based SAR image simulator PolSARproSim. Receiver operating characteristics show that whilst ordinary single-channel CCD is unusable when a canopy is present, 3D SAR CCD permits reasonable detection performance. A novel polarimetric filtering algorithm is also proposed to remove contributions from the ground-trunk double-bounce scattering mechanism, which may mask changes on the ground near trees. To enable this kind of polarimetric processing, fully polarimetric data must be acquired and calibrated. Motivated by an interim version of the Ingara airborne imaging radar, which used a pair of helical antennas to acquire circularly polarised data, techniques for the estimation of polarimetric distortion in the circular basis are investigated. It is shown that the standard approach to estimating cross-talk in the linear basis, whereby expressions for the distortion of reflection-symmetric clutter are linearised and solved, cannot be adapted to the circular basis, because the first-order effects of individual cross-talk parameters cannot be distinguished. An alternative approach is proposed that uses ordinary and gridded trihedral corner reflectors, and optionally dihedrals, to iteratively estimate the channel imbalance and cross-talk parameters. Monte Carlo simulations show that the method reliably converges to the true parameter values. Ingara data is calibrated using the method, with broadly consistent parameter estimates obtained across flights. Genuine scene changes may be masked by coherence loss that arises when the bands of spatial frequencies supported by the two passes do not match. Trimming the spatial-frequency bands to their common area of support would remove these uncorrelated contributions, but the bands, and therefore the required trim, depend on the effective collection geometry at each pixel position. The precise dependence on local slope and collection geometry is derived in this thesis. Standard methods of SAR image formation use a flat focal plane and allow only a single global trim, which leads to spatially varying coherence loss when the terrain is undulating. An image-formation algorithm is detailed that exploits the flexibility offered by back-projection not only to focus the image onto a surface matched to the scene topography but also to allow spatially adaptive trimming. Improved coherence is demonstrated in simulation and using data from two airborne radar systems.en
dc.subjectsynthetic aperture radaren
dc.subjectchange detectionen
dc.subjectpolarimetric calibrationen
dc.titleCoherent Change Detection Under a Forest Canopyen
dc.contributor.schoolSchool of Electrical and Electronic Engineeringen
dc.provenanceThis electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at:
dc.description.dissertationThesis (Ph.D.) -- University of Adelaide, School of Electrical & Electronic Engineering, 2020en
Appears in Collections:Research Theses

Files in This Item:
File Description SizeFormat 
Pincus2020_PhD.pdf25.26 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.