Extraction of power line maps from millimeter-wave polarimetric SAR images

Radar backscatter of power lines has lower values than those of the surrounding ground clutter when the power line is oriented at an off-normal direction with respect to the radar line of sight. For power lines, the traditional detection algorithms that are commonly based on the statistics of the backscatter power of the clutter and target result in excessive false-alarm rates due to very low signal-to-clutter ratio. In this paper, the application of a statistical polarimetric detection algorithm that significantly improves the signal-to-clutter ratio is demonstrated. The coherence between the co- and cross-polarized backscatter components is used as the detection parameter. This statistical detection parameter can be applied to any extended targets such as a suspended cable in clutter background. Detection criteria based on clutter backscattering coefficients, power line size, and aspect angle, as well as the number of independent samples are obtained. The performance of the algorithm for mapping power lines in SAR images is demonstrated using a number of low-grazing incidence polarimetric SAR images at 35 GHz.