The Prediction of Geometric Distortions in Airborne Synthetic Aperture Radar Imagery from Autofocus Measurements

The synthetic aperture radar (SAR) is an imaging system that achieves high azimuthal resolution by tracking individual point scatterers using their phase histories, with the expected phase history of a particular point scatterer being derived from the assumed motion of the airborne SAR platform. Normally, it is assumed that the platform travels along a straight line path and that the radar pulses are emitted at equal spatial intervals. However, the aircraft carrying the SAR will be susceptible to extraneous across tracl motions and errors in the pulse spacing also may occur. These errors manifest themselves. as two forms of image degradation in the final image. Firstly, the image will be defocused due to an error in the expected quadratic phase history, and secondly, the image will contain geometric distortions due to an error in the linear phase history. An autofocus technique can be used to focus the image and produce a measure of the quadratic phase error that in turn can be used to estimate the geometric distortions that will be present in the final image. These distortions can be independently measured by direct comparison with a map of the imaged area. This paper describes the application of these methods to some real SAR data and discusses the results of the comparison of the measurements of autofocus and geometric distortions in terms of both the likely platform motions present and the viability of predicting geometric distortion using the autofocus measurements.     

Comments on “Statistical and spatial properties of forestclutter measured with polarimetric synthetic aperture radar”

For original paper see ibid., vol.30, no.3, p.578-88 (1992). Sheen and Johnston have derived a theoretical expression for the intensity autocorrelation function of a synthetic aperture radar (SAR) image based on the product model for SAR clutter. This derivation is valid under the assumption that the instrument function width is narrow relative to the width of the texture autocorrelation function. This communication draws attention to a more general derivation of the intensity autocorrelation function, given by Oliver, which does not require this assumption     

Optimum classification of non-Gaussian processes using neuralnetworks

A prerequisite for target detection in synthetic aperture radar and moving target imaging radars is an ability to classify background clutter in an optimal manner. Such radar clutter can frequently be modelled as a correlated nonGaussian process with, for example, Weibull or K statistics. Maximum likelihood (ML) provides an optimum classification scheme but cannot always be formulated when correlations are present. In such circumstances, nonlinear, adaptive filters are required which can learn to classify the clutter types: a role to which neural networks are particularly suited. The authors investigate how closely neural networks can approach optimum classification. To this end, a factorisation technique is presented which aids convergence to the best possible solution obtainable from the training data. The performances of factorised networks are compared with the ML performance and the performances of various intuitive and approximate classification schemes when applied to uncorrelated K distributed images. Furthermore, preliminary results are presented for the classification of correlated processes. It is seen that factorised neural networks can produce an accurate numerical approximation to the ML solution and will thus be of great benefit in radar clutter classification     

Estimating building dimensions from synthetic aperture radar image sequences

The problem of automatically extracting building dimensions from synthetic aperture radar (SAR) image sequences of urban scenes is considered. An algorithm based on the delineation of shadows using active contours constrained by a simple wire-frame building model has been developed and demonstrated using SAR imagery of a village on Salisbury plain. The core of the algorithm is a novel technique for target delineation involving multiple active contours evolving simultaneously. In particular, a technique referred to as multiple hypothesis delineation, in which contours can be in several states simultaneously, is developed and shown to lead to considerable improvement in convergence time and delineation accuracy when used to delineate multiple targets in close proximity. The technique is applied to the automatic estimation of building dimensions by delineation of shadows in a sequence of SAR images of an urban scene. The estimation of building dimensions is automatic; user interaction is limited identifying a building of interest and a region of background clutter close to the building. Results are presented for six different buildings, in each case two SAR images were used in the estimation process separated in illumination angle by either 28deg or 90deg. The estimates of building dimensions are compared with the actual building dimensions obtained from architectural drawings. The algorithm was found to perform robustly and provide reasonably accurate estimates of the building dimensions, typically within ~10% of the true values.