Nonlinear Least-Squares Post-Processing for Compressive Radar Imaging of a Rotating Target

In compressive radar imaging of rotating targets, conventional sparse reconstruction algorithms produce blurred and low-contrast reconstructed images of the target due to dictionary mismatch caused by off-grid scatterers. A nonlinear least-squares post-processing (NLSPP) method has recently been developed to tackle the blurring problem existing in the reconstructed images based on cluster analysis and nonlinear leastsquares estimation (NLSE). In this paper, we propose a new improvement of the NLSPP method, which is called the I-NLSPP method, based on a reformulation of the NLSE process of the NLSPP method. Specifically, by jointly performing the NLSE process over all atom clusters using the original backscattered signal, the proposed 1- NLSPP method results in more accurate estimates of the positions and reflectivities of the scatterers constituting the target. The superior performance of the proposed I-NLSPP method over the NLSPP method is demonstrated by way of simulation. In particular, we observe that the 1-NLSPP method achieves a mean-squared-error performance much closer to the Cramér-Rao lower bound than the NLSPP method.