Research Radar Data Sparse Bayesian Learning Algorithm Based On Fast Integration

Since the resolution of radar is based on the bandwidth and the coherent accumulation angle of the signal, multi-radar data fusion has been widely applied in the area of military as a new technique in recent years. Multi-radar data fusion utilize the signal processing technique to obtain high precision parameters of the model of return signal, which can fuse the radar data from different aspect angles and different frequency bands. It breaks restriction of the resolution of the single radar, so it can obtain higher resolution radar imaging. This thesis mainly presents the multi-radar data fusion technique for radar working in high frequency region. By using the geometry theory of diffraction (GTD) model, the radar data fusion problem can be regarded as a signal sparse representation problem. Signal sparse representation technique can be used to estimate the parameters of the GTD model effectively. It can get high quality image. This thesis is mainly divided into four parts:The first part mainly introduces the theories of multi-radar data fusion technique, including the target electromagnetic scattering model and the theory of signal sparse representation.The second part mainly introduces the multi-radar data fusion technique where the radar data come from the same aspect angle and different frequency bands. Firstly we build the signal sparse representation model for the one-dimensional radar target electromagnetic scattering model. Then we analyze the two bands of different radars. We choose the sparse Bayesian learning (SBL) method to resolve the signal sparse representation problem. Three methods including expectation-maximization (EM) method, analytical method and fast marginal likelihood maximization method are introduced to obtain the hyper-parameters.The third part mainly introduces a coherent compensation method based on the signal sparse representation theory. Combining with the signal sparse representation theory in the first part and the sparsity of compensation parameter, we use the SBL method to estimate the linear phase offset and fixed phase offset. Higher precision and better robustness can be achieved by this method.The fourth part mainly introduces the multi-radar data fusion technique where the radar data come from different aspect angles and different frequency bands. Firstly we show the two-dimensional radar target electromagnetic scattering model. Then we build the signal sparse representation model. Finally we use SBL method to resolve the problem and give the simulation results to verify the correctness of the method.