Research Of Compressive Sensing Theory In Radar Imaging

As an important development direction of SAR imaging system, high resolutionimaging radar system has found wide application. The Nyquist sampling theorem basedsignal processing methods have made high resolution radar system confronted manyproblems such as high sampling rate, huge data volume, huge data transmitting, storageand high speed processing. To find new data sampling and processing methods hasbecome an emergent demand.The new concept of Compressive Sensing(CS) is a new signal processing theorydealing with sparse or compressible signals. Motivated by the merits of CS based radarimaging system, such as simplify the hardware of radar system, improve the imageresolution, not sensitive to data loss, many learners have been devoted to the research ofCS based imaging systems. But as a new area, the CS based radar imaging area is just atthe beginning, many problems are needed to be solved. This dissertation concerned onthe basic theory of CS and the application in radar imaging. The problems such as theoptimization of sampling matrix, the problems in the CS based imaging methods, theCS based moving target imaging and the CS based SAR image compression arethoroughly researched. The main works and contributions are as follow:In the aspect of CS theory, a new optimization method of sampling matrix based onequiangular tight frame design is proposed. As equiangular tight frames have minimalcoherence, the optimization method based on equiangular tight frame design candecrease the relevance of sparse matrix and sampling matrix, and results in theimprovement of the reconstruction stability and precision and the area extension of CSapplication.The application of CS based radar imaging methods is thoroughly researched.Sparse representation models of radar echo signal under the mode of de-chirp and matchfiltering is constructed based on the analysis of signal property. A low speed periodnon-uniform sampling A/D is proposed for random real time sampling in range. Afrequency dependent GTD mode is proposed in range imaging which can not only getthe reflectivity but also the information relevant to the scattering kinds. The CS theory is applied in sparse aperture radar imaging in order to improve the stability to data loss.The CS theory is used in spot-mode SAR imaging using Bayesian method which canimprove the stability to noise.The CS based method of joint velocity-DOA estimation and imaging of movingtargets is also researched. In the CS based method, the joint velocity-DOA estimationcan be carried out easily as transforms the problem as a multiple measurement vector.The imaging problem of moving targets is converted to sparsity confined estimation ofmoving parameters and many targets with different parameters can be dealt withsimultaneously. The method can get not only the moving parameters but also theoriginal position, and targets with different parameters in the same pixel can also bedifferentiated.At last, the CS based SAR image compression methods are researched, includingBayesian Matching Pursuit method and K-SVD dictionary learning based method. TheBayesian Matching Pursuit method adopts a hierarchical CS mode in order to constructthe image in the sense of MMSE. In the K-SVD dictionary learning based method, abetter sparse dictionary is constructed based on the image itself which can get a moresparse representation.In all, in this dissertation, the CS theory and its applications have been thoroughlyresearched. The CS based methods are sanguine to solve the problems in high resolutionradar, such as huge data volume sampling, transmission and storage. Still the CS basedmethods can simplify the systems, decrease the demand of high speed A/D and offer thechance for the cancellation of matching filter.