Research On Passive Millimeter-wave Imaging With Irregular Sampling And Image Reconstructing Algorithm With Sparsity

Passive millimeter wave imaging technology makes use of the radiation difference between targets and background in millimeter-band. It can be used widely in the areas of super-low-altitude anti-stealth, surface monitoring and security inspection and so on. While the traditional passive millimeter wave imaging is confronted with many problems such as technology, hardware complexity and cost, which is difficult to use in practice.In the application of super-low-altitude anti-stealth, surface monitoring and security inspection, many areas doesn’t include strong scattering points, and passive millimeter wave image spectrum focus on the low frequency, which shows a typical sparse feature. Therefore, passive millimeter wave imaging technology with irregular sampling provides a solution. Aimed at research on the irregular sampling technologies and image reconstruction algorithms with the sparse priors, the thesis will carries out the following work.1. We research the basic theory of irregular sampling and passive millimeter wave imaging, provided a mathematical model based on passive millimeter wave imaging with irregular sampling.2. We research the block irregular sampling theory, and analyze the reasons caused by the block effect of reconstructing image while adopting the block reconstruction model.Then we introduce ordering operator into the block reconstruction model, and redesign measurement matrix.After that we bring out the global reconstruction model, which can effectively eliminate the block effect.3. On the issues of poor quality reconstructed image for 0-norm minimization, firstly, we research a fixed point iterative algorithm, then come up with the fixed point continuation algorithm based on mixed norm after introducing continuation strategy. We also analysis and present a proof for the convergence of the fixed point continuation algorithm.Compared with the fixed point iterative algorithm, the fixed point continuation algorithm reailises the improvement of performance.4. Contraposing the problem that the fixed point continuation algorithm can’t solve the large sparse problem which is not sparse enough, we introduce non-monotone line search method and subspace optimization into it, and give the active-set sparsereconstruction algorithms for 1-norm minimization.Finally, the simulation results show that compared to block reconstruction, the algorithm based on block sampling and global reconstruction model effectively eliminate block effect in reconstructing images, and the reconstruction performance based on the fixed point continuation algorithmis is better than other reconstruction algorithms for 0-norm minimization, and the quality of the reconstruction images obtained by the active-set sparse reconstruction algorithms is better than by the previous two algorithms.