چكيده به لاتين
In ISAR imaging, it is very important to have a high resolution in order to have a clear and accurate image to detect targets. Due to the two-dimensional nature of ISAR images, the resolution in the range direction depends on the bandwidth of the transmitted signal and the resolution in the azimuth direction depends on the coherent processing interval, which due to the existing limitations, is not always possible to increase the resolution in any direction. The main goal of this thesis is super-resolution in ISAR imaging of the data obtained from a limited number of observations using the Gridless method and increasing the two-dimensional resolution in order to increase the accuracy in detecting the nature of the target in ISAR images. There are various methods of super-resolution, including methods based on spectrum estimation, SBL methods, grid-based sparse methods, and gridless-based sparse methods. As the first proposed method, the FRWTM method by defining two-dimensional weighted atoms and defining the atomic norm, the atomic norm minimization problem, which is a convex optimization problem, is transformed into the Trace minimization problem of a two-level Toeplitz matrix and by Vandermund decomposition, the scattering points of the target is recovered and ADMM method is used to increase the speed. In the performed simulations, the target image can be recovered by having only 15% of the samples and in another simulation. In the next proposed method, the goal is super-resolution in ISAR imaging of space objects such as space debris and satellites. In ISAR imaging of space debris due to their rapid rotation, it is necessary to have enough snapshots to fully recover the image. To solve this problem and improve the quality of recovered images, several snapshots are used, and to reduce the amount of data, calculations, and processing time in each snapshot, randomly selected spot beams are used in each snapshot, which is called Encoded Aperture. To determine the active/inactive spot beams in each aperture coded from the Bernoulli distribution, and to recover the image, L1 and TV norms have been used. The simulation results show the superiority of using the L1 norm in the first scenario and in recovering ISAR images of space debris in the number of 100 snapshots, while in the second and third scenarios, the TV norm in image recovery in the number of snapshots Less, i.e., 100 snapshots in signal to noise of 5 dB is more successful. Another proposed method is FRAND, which is based on the definition of two-dimensional atoms, atomic norm minimization and Noise reduction, Toeplitz matrix, vandermonde decomposition, and a new weighting matrix, in order to increase the speed of the proposed method, the ADMM method is used. In the performed simulations, the proposed FRAND method has succeeded in recovering the target image with a limited number of available samples. All the proposed methods have been compared in terms of MSE with other related methods in different signals to noise and the superiority of the proposed method has been proven.
