Research On Compressed Sensing Algorithm Applying To UWB Radar

Ultra-wideband radar obtains wide utilization because of long-distance high resolution, low probability of intercept, anti-jamming, low power dissipation, anti-stealth, and other advantages that the narrow-band radar cannot meet. Compressed sensing(CS) originate in the fuctional analysis and approximation theory. This theory breaks through the restriction of Nyquist sampling theorem and makes the sampling rate not depend on the signal bandwidth, but on valuable information in the structure and content of the signal.Due to ultra-wide bandwidth and high frequency of UWB radar, according to the Nyquist sampling theorem and the limination of uncertainty principle of radar, the front-end sampling system and back-end image storage systems confront greater pressure. Therefore, regarding the applications of life detection and ultra-wideband synthetic aperture radar system of ultra-wide bandwith radar, this dissertation proposes a compressed-sensing-based sampling algorithm, image coding and image fusion.Specific work in this dissertation is as follows:(1) Compressed sensing theory consists of signal sparsification, under-sampling measurement and recovery algorithms. It could achieve the qualified sparse signal observations by measuring frequency of far less than the length of the signal. The range of sampling rate is determined by the minimum number of measurements. This dissertation explores the mathematical reasoning method of smallest number of measurement in compressed sensing reconstruction algorithm, and be vertified and applied in experiments and simulations that follow.(2) Ultra-wideband life detection radar based on the principle of micro-Doppler, could detect the body’s breathing, heart rate and other vital signs. Regarding long time smapling and the problem of handling large volumes of data, by radar echo signal analysis, this dissertation sets up a modle on life body respiratory signals detection with stepped continuous wave UWB radar and uses compressed sensing approach to sample respiratory data. This dissertation designs self-adaptive compressed-sensing-based respiratory signal sampling algorithm and uses distance-domain filtering method to eliminate noises and interferences to achieve respiratory frequency detectio n. Through experimental verification, respiratory frequency could be perfectly detected, realizing life radar undersampling when sampling rate is 0.66.(3) For the problem that the wall- through radar images are discrete into grid when compressing bringing huge amount of data and problem of possible target matrix sparcification in block, this dissertation takes UWB radar images which use complex FastICA algorithm method to eliminate direct signal as research object, seeks efficient and robust CS compressing algorithm, proposes CS radar coding algorithm to improve the performance of large-scale reconstruction of the image. Because the realization of the observation matrix is the major obstacle in the process of hardware realization, by the simple and efficient screening and pretreatment of measurement matrix, the algorithm weaken correlation between measurement matrix elements, thereby effectively improving the image quality. Firstly, through the design of measuring sparse matrix and simulation of the convex optimization algorithm and the greedy algorithms, herein the applicable block sparse radar image coding algorithm is determined. The use of linear prediction algorithm based on minimal residual, could obtain accurate position and a clear outline of the target image in the case of lower compressive sampling rate and reduce of the number of required imaging probe and save time resources.(4) For growth of synthetic aperture radar imaging area, increasing demand for higher radar imaging resolution and signal bandwith and increasing difficulty of signal acquisition and processing, this dissertation proposes orthogonal- matching-pursuit-based two wavelet fusion algorithms. For large ultra-wideband SAR fused image, in view of the wavelet sparse characteristic feature of radar images, after wavelet transformation, firstly, set large wavelet coefficients of high and low frequencies into packets respectively. According to the least-squares criterion, determine the relative error of each packet, record the position index of the non-zero element of sparse solution. According to the recorded position index of non-zero elements, build the least squares method; fuse each group of wavelet coefficients in pixel level; find the most suitable sparse wavelet base and fusion bases for radar images; finally, reconstruct with OMP recovery algorithm and compare with the SAR image fusion algorithm of GPSR method. Ultra-sparse fusion algorithm proposed in this dissertation greatly minimizes the amount of processing data, realizes ultra wide band high frequency SAR image fusion, and effectively avoid the blind pursuit of high-cost problems caused by the high-resolution detector.