Study On Microwave Radiation Image Reconstruction Method Based On Sparse Representation

Microwave remote sensing has the characteristics of all-day, all-weather, strongpenetrability and concealment; it is widely used in the remote sensing, radio astronomy,disaster forecast, and military target detection, etc. Microwave radiometer reflects the targetintrinsic physical characteristics information by measuring the target microwave radiationcharacteristics, but it is not sensitive to surface roughness macrostructure characteristics of thetarget. The characteristics obtained from soil microwave radiation image using microwaveradiometer mainly depend on soil moisture. We can get soil moisture data by inverting theobtained microwave radiation image, and improve the accuracy of weather forecast byanalyzing the soil moisture data, and effectively monitor geological disasters, such as drought,flood, etc. With the development of refinement and structuralization of the image,interferometric synthetic aperture microwave radiometry (ISAMR) has evolved into anenormous and complex system for the demand of high resolution, and it can easily reach tensof millions when collecting data in one snapshot. So it is difficult to achieve the highresolution microwave radiation imaging based on the Nyquist sampling and conventionalmicrowave radiation imaging method.Combined with the sparse prior knowledge of signal for signal reconstruction,compressed sensing (CS) utilizes the random sparse sampling (far less than the Nyquistsampling rate) to compress data, and adopts the nonlinear reconstruction algorithm toreconstruct the original signal. In this paper, we apply CS to microwave radiation imaging,reconstruct the sparse sampling data by reconstruction algorithm, and obtain the highresolution microwave radiation image. The main research content of this paper as follows:1. Introduce the interferometric synthetic aperture microwave radiation imaging system;analyze its advantages and disadvantages. Discuss the theoretical framework of CS, brieflyintroduce the observation matrix; the main research is sparse representation of signal and thesignal reconstruction algorithm, construct the model of microwave radiation imaging based onCS.2. Study microwave radiation image reconstruction method based on orthogonalmatching pursuit (OMP) algorithm and optimized orthogonal matching pursuit (OOMP)algorithm. Introduce the Gaussian random observation matrix, and verify the effectiveness byexperimental simulation. Theoretically analyze the structural characteristics of the microwaveradiation image, and obtain the prior knowledge of microwave radiation image, such aspiecewise smoothness, sparsity in transformation domain. Introduce the combined dictionary to OMP algorithm, experimental results show that combined dictionary obtains higherresolution microwave radiation image than single orthogonal basis. Study the microwaveradiation image reconstruction method based on OOMP algorithm, experimental results showthat the quality of reconstructed image and the reconstruction precision by OOMP algorithmis better than that by OMP algorithm.3. Introduce K-SVD dictionary learning algorithm, take example by present popularDLMRI algorithm and GradDLRec algorithm, study the adaptive multi-structural sparsedictionary, and improve the sparsity of the image and the adaptability of the dictionary.Introduce Bregman iterative algorithm, design microwave radiation image reconstructionmethod based on adaptive multi-structural dictionary learning, experiments verify theeffectiveness of the proposed algorithm.