Mark Preiss

Detecting scene changes using synthetic aperture Radar interferometry

In repeat-pass interferometric synthetic aperture radar (SAR), man-made scene disturbances are commonly detected by identifying changes in the mean backscatter power of the scene or by identifying regions of low coherence. Change statistics such as the sample mean backscatter-power ratio and the sample coherence, however, are susceptible to high false-alarm rates unless the change in the mean backscatter power is large or there is sufficient contrast in scene coherence between the changed and unchanged regions of the image pair. Furthermore, as the sample mean backscatter-power ratio and sample coherence measure different properties of a SAR image pair, both change statistics need to be considered to properly characterize scene changes. In this paper, models describing the changed and unchanged regions of a scene are postulated, and the detection problem is expressed in a Bayesian hypothesis-testing framework. Forming the log-likelihood ratio gives a single sufficient statistic, encoding changes in both the coherence and the mean backscatter power, for discriminating between the unchanged- and changed-scene models. The theoretical detection performance of the change statistic is derived and shows a significant improvement over both the sample mean backscatter-power ratio and sample coherence change statistics. Finally, the superior detection performance of the log-likelihood change statistic is demonstrated using experimental data collected using the Defence Science and Technology Organisations Ingara X-band airborne SAR

Polarimetric phase gradient autofocus

In the past decade the use of fully polarimetric SAR (polSAR) systems has increased significantly due to their effectiveness in target classification and detection applications. While polSAR imagery has been extensively used to distinguish between different scattering mechanisms in a scene, there has been a lack of research in the exploitation of polarimetry to assist in image formation and in particular autofocus for fine resolution SAR. In this paper an extension of the phase gradient algorithm (PGA) for polSAR imaging is proposed and its effectiveness is tested on simulated and real data.

Image formation on undulating terrain using the upgraded Ingara L-band radar system

A beamforming approach to SAR image formation is shown to facilitate spatially variant aperture trimming which maximises the potential repeat-pass coherence when focusing onto undulating terrain. The method is demonstrated using the upgraded Australian airborne radar system, Ingara.

Quad-polarimetry and interferometry from repeat-pass dual-polarimetric SAR imagery

New space-borne polarimetric SAR systems employ, or will employ, dual-pol imaging modes as well as full quad-pol imaging, e.g. PALSAR, RADARSAT-2, TERRASAR-X. Therefore questions arise as to the capabilities (and limitations) of these various dual-pol SAR imaging modes. One novel offshoot of our work on dual-pol SAR image analysis and polarimetric decompositions of dual-pol data was the idea to collect two dual-pol images employing geometry appropriate for parallel repeat-pass interferometry. The polarizations of the dual-pol collections differ, e.g. one is (HH, HV) and the other (VV, VH). Using the example of transmitting either horizontal (H) or vertical (V) polarizations, one collects all four polarimetric channels, but only two per pass. Except for the interferometric repeat-pass nature of the SAR image collections the polarimetric information should be identical to standard quad-pol SAR imagery. The relevant open questions concern the interferometric baseline and temporal decorrelations between the repeat-pass dual-pol images.

Physics-based predictions for coherent change detection using X-band synthetic aperture radar

A theoretical model is developed to describe the interferometric coherency between pairs of SAR images of rough soil surfaces. The model is derived using a dyadic form for surface reflectivity in the Kirchhoff approximation. This permits the combination of Kirchhoff theory and spotlight synthetic aperture radar (SAR) image formation theory. The resulting model is used to describe the interferometric coherency between pairs of SAR images of rough soil surfaces. The theoretical model is applied to SAR images formed before and after surface changes observed by a repeat-pass SAR system. The change in surface associated with a tyre track following vehicle passage is modelled and SAR coherency estimates are obtained. Predicted coherency distributions for both the change and no-change scenarios are used to estimate receiver operator curves for the detection of the changes using a high-resolution, X-band SAR system.

3D SAR beamforming under a foliage canopy from a single pass

Coherent detection of changes on the ground under a forest canopy by repeat-pass synthetic aperture radar (SAR) imaging is problematic due to the mixture of ground and canopy responses. 3D SAR imaging, by beamforming data from multiple low-frequency across-track acquisitions, offers a way to separate the forest components in height. However, data acquired from multiple passes require precise registration of the flight tracks, which is often not possible, particularly for small airborne platforms. This study analyses the potential for SAR imaging of the ground under a forest canopy using data from a single pass of a multichannel across-track radar system. We focus in particular on the case of a two-channel alternating-transmit interferometer giving three effective input channels. 3D image formation in one pass by phase-preserving adaptive beamforming is shown to provide sufficient attenuation of the interference from a model forest volume to permit a reasonable estimation of the ground coherence across two passes for coherent change detection.

Aspects of 3D tomography for multiple-pass spotlight-mode airborne SAR

Three-dimensional SAR tomography makes use of radar data acquired at different depression angles to synthesise an aperture in elevation and thereby undo layover and resolve along height. Focusing on one tomographic method appropriate to a spotlight-mode airborne collection system, the details of spatial bandwidth usage and point-spread function are described, and the phase modulation effects of the initial two-dimensional image formation stage are analysed in terms of the impact on the tomographic processing. Finally, an adaptive beamforming approach is investigated for the improvement of height resolution.