SAR image classification by image intensity similarity and kernel method

摘要

Aiming at the classification problem of Synthetic aperture radar (SAR) images, a classifier based on image intensity and structure is constructed. To overcome the disadvantages of conventional template matching algorithms, similarity between two images is calculated by Hausdorff function, which can handle the distortions and pixel perturbations. The function is then fed into Support vector machines to eventually accomplish the task of image classification. Experiment results corresponding to field and simulated data show that this method characterizes target structure information well.