Research On Radar High Resolutionimaging Technologies Based On Compressive Sensing

Radar high resolution imaging technology is an important research topic. It has been applied in many fields, such as scene observation, target feature analysis and target recognition. Imaging resolution is an important factor for evaluating radar imaging quality. However, higher resolution relies on a radar system with higher performance, which causes the signal processing method based on the Nyquist-Shannon sampling theorem suffer from more troubles, such as high-speed data collection, storage and transmission of large amount of data, rapid processing of large amount of data. These problems raise great challenge to the hardware system and resources, and limit further improvement of radar resolution. Therefore, it is necessary to find new data acquisition and imaging methods for radar imaging development.The recently proposed Compressive Sensing(CS) theory provides a novel approach for radar imaging technology. It is not restricted by the Nyquist-Shannon sampling theorem, and breaks through the limitation of time and frequency uncertainty principle. It provides higher resolution for radar imaging. This dissertation studies CS high resolution imaging technologies based on the basic theory of CS and its application in radar imaging, i.e., moving target CS high resolution profiling, CS high resolution profiling in noise level unknown case and Inverse Synthetic Aperture Radar(ISAR) CS high resolution imaging under low signal to noise ratio(SNR) situation. The main work is summarized as follows.1. The velocity compensation and high resolution profiling problem of moving target based on CS is studied. The profiling quality of moving target greatly relies on the velocity compensation precision. The influence of target motion on high resolution profiling of step frequency waveform is analyzed, and an accurate velocity compensation method compound time domain correlation and CS is proposed. This method uses two adjacent frame high resolution range profile(HRRP) to achieve speed coarse compensation by time domain correlation processing, and conductes velocity compensation on these two frame return data. Then the two frame data are conjugately multiplied, and a center frequency item is yielded in the result, which has the largest amplitude. A redundant dictionary is established to sparsely represent the center frequency item. By solving the constrained optimization problem with sparse degree 1, the accurate velocity estimation can be obtained for accurate velocity compensation. Meanwhile, moving target high resolution range profiling based on random frequency waveform is studied, and the simulation experiments are provided to verify the performance.2. CS high resolution profiling under noise level unknown situation is studied. The profiling quality of CS method under noise circumstance is impacted by noise level, which however is usually unknown. Based on the model of high resolution profiling by CS in noise circumstance, a CS signal reconstruction method based on noise level estimation by sliding sub-sequence processing is proposed, which is suitable for single wideband pulse echo. The sampling data is divided into sub-sequences, and a sub-sequence correlation matrix is established to estimate the noise level. The noise level is used for CS signal reconstruction, and an accurate high resolution profile is achieved.3. ISAR imaging method based on CS under low SNR is studied. ISAR imaging at low SNR suffers from noise pollution. To improve imaging quality, a scattering area weighted CS imaging algorithm is presented. This method uses the target scattering area information to weight the basis function in redundant dictionary, and revised CS reconstruction algorithm to suppress noise pollution and improve the imaging quality. The imaging performance of this method is verified by simulation and real data experiments.4. The actual application of CS fast reconstruction algorithm in ISAR imaging in low SNR circumstance is studied. ISAR imaging demands real-time processing. However, CS fast reconstruction algorithms usually have noise sensitivity. This results in the imaging results seriously impacted by noise. To deal with the problem, a sub-sequence singular value decomposition(SVD) approximation method is proposed. This method focuses on improving the echo SNR, also gives attention to imaging under short CPI. It establishes the sub-sequence matrix by exploiting the echo signal sparse characteristic. The noise components in the principal components are removed through SVD, and the denoised signal is reconstructed by these principal components. With the denoised signal, the high quality image is achieved. The performance of this method is evaluated by simulation and real data experiments.