<p>High-resolution inverse synthetic aperture radar (ISAR) imaging with large rotation angle can be accomplished by applying terahertz (THz) radar for space target detection. Nevertheless, due to the non-cooperative characteristic of the target, the unknown attitude transformation and rotation parameters pose a challenge in accumulating energy and matching angular information for high-resolution imaging with large rotation angle. Moreover, in the terahertz band, the scattering characteristics of space target with complex structures are complex and hard to divide sub-apertures accurately. To address the abovementioned challenges, in this paper, we employ the energy accumulation curve to fast divide the large-angle echo data into several sub-apertures, thereby eliminating the influence of target attitude transformation or maneuvering on imaging accumulation. Subsequently, for the sub-aperture echo data with relatively stable motion state, a convolution back projection (CBP) imaging algorithm based on modified particle swarm optimization (MPSO) is adopted. The rotation parameters are accurately estimated under the criterion of maximum image contrast, consequently achieving accurate matching of angular information and high-resolution imaging of the target. Simulation and experimental data of terahertz radar fully demonstrate the correctness and effectiveness of the proposed algorithm, achieving accurate estimation of the rotational parameters and high-resolution imaging of non-ideal rotating target.</p>