AUTOMATIC DETECTION OF TARGETS AGAINST CLUTTERED BACKGROUNDS USING A FRACTAL-ORIENTED STATISTICAL-ANALYSIS AND RADON-TRANSFORM

The paper concerns a two-stage method of local and global information processing used to extract low contrast, linear features embedded in SAR images degraded by speckle. The first processing stage which involves the scaling laws of fractal geometry identifies regions in images where there is a high probability of a target feature. Such features cause deviations from the statistics for a background only. Thus by computing these deviations, the statistics of the target features are automatically separated from those of the background. The result of this processing is then passed, in the form of a binary image, to the second stage where it is parametrically transformed and the thresholded transformation is used to deduce the location of the target feature within the image. The method is robust, automatic and can be shown to be computationally realistic. It is illustrated using data from a SAR image (C-band, VV polarisation).