High-resolution radar imaging systems employ wideband signals and large array apertures. As a result, large amounts of data are generated, which presents challenges in both large amounts of data and data processing. Therefore, the development of computationally efficient algorithms and the reduction of the number of spatial and frequency measurements are highly desirable.In recent years, low rank matrix reconstruction theory becomes the hot topic in the image signal processing. The low rank matrix reconstruction problem can be solved by the nuclear norm regularized linear minimum mean square optimization problem if the signal is low rank or approximation low rank. Low rank matrix reconstruction theory brings a good chance for the radar imaging system.In this paper, low rank representation is applied to the radar imaging systems. Firstly, Numerical analysis is carried on with accelerated proximal gradient singular value thresholding algorithm. Compared with the compressive sensing technology, the accelerated proximal gradient singular value thresholding algorithm for the reconfiguration problems in the radar imaging is a higher computing speed, resolution and robust. Secondly, the low rank representation is applied to quickly and precisely reconstruct the low rank data matrix in the electromagnetic inverse scattering imaging field. Simulation results and analysis are carried on with the low rank representation and Linear Born approximation. Compared with the truncated singular value decomposition, the novel algorithm has good robustness. Finally, the reconstruction results based on the measured data shows that the low rank constraint method has good robustness.Both the simulation result and the experimental data show that: firstly, the ground penetrating radar imaging based on the low rank representation is good for committing the target location and shape. Secondly, the method of low rank representation and linear born approximation has the advantage of high space resolution, accurate orientation, fast imaging and so on. The novel algorithm has strong robustness for image processing than the truncated singular value decomposition algorithm in radar imaging systems. Finally, the experimental data show that the low rank representation can quickly and precisely reconstruct the one, two and three filled dielectric cylinders’ shape and relative location. |