Tensor Compressive Sensing Based Staring Imaging Using Multiple Transmitter Radar

The traditional staring imaging radar utilized real aperture imaging techniques. By illuminating the scene of interest by a narrow beam via beam forming, we can reconstruct the image of the scene with received echoes. The resolution is one of the most important criterions to measure the performance of radar imaging. According to the Rayleigh criterion, the method of improving the azimuth resolution is to increase the antenna aperture and the frequency of the transmitted signal. However, the expansion of the antenna aperture will cause the increase of the antennas' size and weight, which brings great difficulty to the design and application of the antenna. The main contributions of this thesis are summarized as the following two aspects: 1. Multiple-transmitter staring imaging radar transmits random signals to form random radiation field on the scene. It can break the Rayleigh diffraction limitation by correlating the scattered echo data with the temporal-spatial random radiation field. By investigating the system model of the compressive sensing based multiple-transmitter staring imaging radar, we derive the expressions for signals of the radiation field and the received echoes. The comparison between the calculated radiation field and experimental measurements verifies the theoretical analysis. The imaging result is obtained by utilizing the compressive sensing algorithm, which proves that the imaging method can not only improve the resolution, but also reduce the data storage and computation load. The effectiveness of the imaging method is verified. 2. The conventional compressive sensing based staring imaging method is typically performed by converting the two-dimension matrix to a single long vector, then compressive sensing method is used to reconstruct the image. By converting the vector to the original two-dimensional matrix, we obtain the image of the scene. Application of this method to big scene will cause the sampling matrix to be very large, thus imposing a huge computational and memory burden. This thesis mainly deals with the problem of high computational complexity and long time operation problem in the actual applications. We extend single receiver system model to multiple receiver system model, based on which a two-dimensional signal model using the second-order tensor representation is proposed. A new method of staring imaging based on tensor compressive sensing theory is proposed in this thesis. We demonstrate in numerical experiments that the algorithm can greatly reduce the complexity of the algorithm as well as the data storage burden. The tensor compressive sensing algorithm can provide super fast computations, even for a large scale problem without sparsity constraint.