Radar image resolution depends on physical parameters of the imaging system, such as bandwidth and aperture size. Synthetic aperture radar (SAR) and inverse synthetic aperture radar (ISAR) improve resolution by artificially increasing the size of the aperture. The standard method of back projecting the measured radar cross section (RCS) of a target using an inverse Fourier transform is computationally taxing and is limited by the physical parameters just mentioned. This can be circumvented using different signal processing methods or neural network based methods. This thesis takes an approach based on the geometrical theory of diffraction (GTD). GTD uses a physical model that describes the scattering from different geometries. By using GTD and attempting to classify the scattering geometries of the target by summing up the scattering contributions of different point scatterers, the actual geometry of the target can hopefully be predicted. This is the main contribution of this master’s thesis.