Research On The Algorithm Of Passive Millimeter Wave Imaging Based On Compressed Sensing

Under the condition that passive millimeter wave imaging (PMMW) technologyhas many unique advantages，it has widely significantly applied on the area of aircraftlanding, airport or port scenes monitoring, security check in import and export ofbanks, stadiums or other important places, etc. Nevertheless, to meet the requirementsof high-quality image in the field of high-performance scene monitoring, and real-timeimaging in the field of anti-terrorism and security check, the traditional PMMWimaging technology need to increase the complexity of the system hardware and thecost, for example of large-caliber antenna size, a large number of the receive array, etc.However, all of these methods are hard to put into practice.Fortunately, a new developed theory named compressed sensing that can offer arevolutionary solution. Based on the sparsity of signal and sub-Nyquist non-coherentcompressive sampling, one could high precisely reconstruct the original signal throughthe algorithm of sparse optimization. This thesis researches the theory of compressedsensing and its application on the domain of PMMW imaging. The main contents ofthis thesis include:1. We research the theory of PMMW imaging and compressed sensing, includingsparse representation, measurement matrix and the reconstruction algorithm of sparseoptimization. We also analyze the mathematical model of PMMW imaging systembased on compressed sensing, and conceive the design of this system practically.2. According to the reason that the traditional method of l1minimization couldnot preserve the edges or boundaries of image more accurately, we study a newefficient TV minimization scheme based on augmented Lagrangian and alternatingdirection algorithms, short for ALTV scheme. This algorithm introducesvariable-splitting technique, using the alternating direction method to minimize theaugmented Lagrangian function to solve the variable. While either one of two variablesis fixed, minimizing the function with respect to the other has a closed-form formulawith low computational complexity and high numerical stability.3. According to the requirements of which the implementation of matrix-vector multiplication from ALTV has a large of computational complexity, we analyze the fastwalsh hadamard transform which could further speed up ALTV algorithm run rate.4. We research the features of point spread function (PSF) of PMMW image,designing a corresponding variable density measurement matrix. The simulationexperiments verify that compared with the uniform density measurement matrix, it hasa better performance of reconstruction.5. According to how to significantly improve the sparse representation of image,we propose an improved TV+Wavelet model based on TV model, with the introductionof sparse prior information of PMMW image. Then, we adopt split Bregman iterativealgorithm to reconstruct PMMW image.Simulation experiments show that compared with the traditional TV minimizationalgorithm, ALTV algorithm has faster convergence speed and higher performance ofreconstruction. We also prove that compared with TV model, based on improvedTV+Wavelet model, the split Bregman iterative algorithm that reconstructs PMMWimage has better performance.