MVM based SAR image processing for ship pose estimation

As an essential approach of SAR image interpretation, a good pose estimation calls for two conditions: an accurate extraction of the scattering centers and a thorough exploration of the structural information among them. The article focuses on these two topics successively. Firstly, the power spectrum estimation methods are recommended to improve the scattering center resolution and the Minimum Variance Method (MVM) is chosen for its sidelobe suppressing ability and signal model independence. With the segmented MVM image, the axis extraction problem is then solved by a new strategy, namely the Angle Entropy of Radon transform (AER) strategy. Data from computational electromagnetics are used for experiments. It is shown that the MVM image shows better morphologic feature than the Fourier ones (both original and Hanning-windowed), and the AER strategy achieves more accurate and robust estimation compared with the edge-based Hough transform (HT) technique.