Range-Doppler Imaging Based On Sparse Representation

A fundamental task of conventional radar is to detect and localize radar targets. In recent years, the application of radar is more and more widely, such as a most important one-radar imaging. In the development of imaging technology, the high resolution imaging is one of the key technologies. Stepped-frequency technology is a very important technology of realizing the high range resolution imaging. Using stepped-frequency technology can construct a wide-bandwidth chirp pulse by combine a burst of narrow-bandwidth chirp pulse transmitted at stepped-frequency intervals. This synthetic wide-bandwidth has a same range resolution as the wide-bandwidth. The advantage of this technology is the reduction of instantaneous bandwidth and sampling rate requirement in the radar system. Sparse representation has drawn much attention from researchers due to its ability of reconstruction the signal from a small amount of measurements in recent years. However, there are still many problems existed in its application on radar range-doppler imaging. This thesis mainly studies the high resolution technology of stepped-frequency signal and range-doppler imaging based on sparse representation.Firstly, this thesis introduces the frequency step signal and its principle of synthesis and then deduces the time domain expression of stepped-frequency signal, which used the linear FM signal as its sub pulse. This thesis focuses on the analysis of two high resolution processing technology of stepped-frequency, namely the time-domain bandwidth synthesis and the frequency-domain bandwidth synthesis. Their advantages and disadvantages are discussed.Secondly, this thesis introduces the traditional range-doppler imaging method, brought in Compressed Sensing, range-doppler imaging question can be regarded as sparse signal recovery problem under the framework of Compressed Sensing. This thesis also establishes a separable range-doppler imaging model based on sparse representation and generalizes one-dimensional SL0 algorithm to two-dimensional. A range-doppler imaging method based on sparse representation is brought in.At last, based on the previous introduction of separable range-doppler imaging model, this thesis applies Bayesian framework to solve the problem of range-doppler imaging.This thesis also introduces the basic theory of Bayesian Compressed Sensing and its fast algorithm. Multi-Task Bayesian Compressed Sensing theory is analyzed. Aiming at the problem of complex parameter estimation, this thesis introduces the noise variance needed to estimation to the model and named this method fixed Multi-Task Bayesian Compressed Sensing algorithm. This thesis also proposes a Multi-Task Bayesian Compressed Sensing algorithm based on separable range-doppler imaging model. The simulation results are given.