Interpolation of ISAR Images Based on Principal Component Analysis with Small Samples

Inverse synthetic aperture radar (ISAR) images that characterize the complete scattering characteristics of enemy aircraft are difficult to be obtained due to the attitude sensitivity of aircraft targets. An interpolation method based on the principal component analysis (PCA) is proposed to achieve the interpolation of ISAR images in this paper. First, an ISAR image pre-processing method is proposed to reduce the effect of ISAR image rotation characteristics. Then, the principle of PCA is analyzed and a "pairwise" principle is proposed based on the analysis of the image features corresponding to the eigenvalues of ISAR images to inverse ISAR images. Finally, a spherical manifold that characterizes ISAR images with different attitudes is established, while interpolation of ISAR images is achieved by interpolation of geodesic lines on the spherical manifold. Furthermore, compared with the image pixel point interpolation, the deep learning method, and the proposed method in this paper, simulation results indicate that the proposed method is promising.