Sparsity-aware Sensing For Space Targets

With the development of science and technology, human space activities become more and more frequent along with constantly expanding scale, which leads to the fierce competition of space resources among the countries all over the world. As one of the most important technologies to dominate the priority of outer space, space target detection and recognition technology has received extensive attentions and relevant studies are of great significance for guaranteeing national and people’s interests and security. Being aware of the characteristics of space targets, such as high velocity, traveling with false targets, complex motion patterns, etc. this thesis focus on investigating four hot topics:high-velocity group target detection, spinning targets imaging, micro-Doppler analysis, precession-with-nutation target modeling and motion parameters estimation. The main contents can be concluded as follows.1. Study on high-speed group space target detection and resolving. For high-speed group targets, it is prone to produce serious range migration and Doppler ambiguity. Besides, there exist two more problems:1) The Doppler ambiguity numbers of the radar returns from a single target may be different at the frequency points within the bandwidth; 2) The tiny speed differences among the group targets may lead to different Doppler ambiguity numbers. All these intrinsic factors cause that the traditional keystone transform fails to detect the high-speed group targets. To overcome the above mentioned shortcomings, we propose a novel keystone transform based on sparse techniques. The new method adopts the return model of high-speed targets. With the assumption that the group targets are sparsely distributed in the range-velocity plane, we reformulate the traditional keystone transform to be a sparse reconstruction problem, and solve it by BLOOMP algorithm. The new approach allows the coherent accumulation detection of high-speed group targets and multi-targets resolving.2. Study on spinning targets imaging. Usually, there exist two problems in spinning targets observation:1) Large time-width signal is widely adopted for space targets detection to get a high signal-noise-ratio, which causes the radar pulse repetition frequency (PRF) can not be high enough and it may lead to the Doppler ambiguity; 2) When the rotation axis of the target does not coincide with the radar line of sight, some parts of the target will be unsighted to the radar, which is referred to as the occlusion effect. However, these two factors are seldom considered in the traditional methods. Firstly, we develop a novel approach for spinning target imaging based on sparse signal reconstruction, referred to as sparse Doppler-only snapshot imaging (SDOSI) method. The basis dictionary of the signal is constructed by the grid model according to the prior moving information of the debris and the spinning target return model. The reconstruction is implemented by an effective greedy method named regularized orthogonal matching pursuit (ROMP). The proposed method is of high imaging accurage and able to image the target when the radar PRF is less than the Doppler bandwidth of the return signal and the occlusion effect with small shadowed size is present. Secondly, we propose a novel approach to image spinning targets with serious occlusion effect based on 2D fused Lasso model, which is referred to as fused Lasso based Doppler-only snapshot imaging (FLDOSI). Due to the spinning motion, the radar illuminates different parts of the target at each observation instant, which causes the backscattering intensities of the scatterers are slowly time-varying. The 2D fused-Lasso model is introduced to capture this time-varying characteristic, and the imaging problem is transformed to be a 2D sparse reconstruction problem. The proposed method is able to adaptively estimate the shadowed region, do Doppler imaging and work well in low PRF situations.3. Study on micro-Doppler analysis of space targets. Since the micro-motion feature is regarded to be able to uniquely capture the motion characteristics of space targets and provide reliable evidences for recognition, how to effectively extract the micro-Doppler feature becomes a key problem. Time-frequency analysis is a great tool to extract the micro-Doppler spectrum of interest. However, traditional time-frequency analysis methods are suffering from the uncertainty principle between the time resolution and frequency resolution or seriously interference terms. To overcome the limitations, we propose to apply the forward-backward time-varying autoregressive (TVAR) model to analyze the micro-Doppler. To get more precise results, sparsity and group sparsity priors are introduced to the model based on the property of radar returns from rigid bodies. The proposed methods are able to well extract the micro-Doppler features from a limited amount of data and fill in consecutive discontinues measurements.4. Study on sensing the micro-motion of space targets based on wideband measurements. Both mathematical modeling and motion parameters estimation are elaborately considered under two assumptions:global scattering center model and slipping scattering center model. The motion parameters estimation includes four steps. Firstly, estimate the high-resolution radial ranges by sparse techniques. Secondly, associate the trajectories of the multiple scatterers by using the dynamic programming method. Thirdly, by using the association results of the histories of the 1D high resolution range measurements, we obtain the Euclidean reconstruction of the motion matrix by factorization based reconstruction algorithm. Finally, we simultaneously estimate the nutation and precession parameters and remove the arbitrary 3D rotation in the Euclidean reconstruction by alternating between the sequential quadratic programming and nonlinear least square optimization. Moreover, sparse reconstruction technique is employed to deal with the case of incomplete 1D radial projections caused by shadowing, specularity, RCS fluctuation, observation failure and other factors.In this dissertation, a lot of experiments based on the numerical simulated data and the electromagnetic (EM) analysis data are carried out to verify the proposed theory and methods. Experimental results demonstrate certain theoretical guidance significance and practical application value of the achievements in sensing space targets.