Multi-Dimensional Moving Target Detection Methods For Array Radar System

Ground moving target indication is one of the most important tasks for surveillance and monitor radar system. Owing to the air-to-ground illumination geometry, moving targets with relatively low velocities would be embedded in the strong clutter echoes, and how to extract the target signal from the compounded data is an important problem for GMTI. Incorporated with the adaptive array processing techniques, space time adaptive processing(STAP) and its extended algorithms are effective to suppress the clutter component in the received data, which are primary approaches that would resolve this problem. Along with the improvement of radar techniques, multi-dimensional systems, such as the multi-frequency or multi-polarization radar systems, are developed or undergoing experiments. In this thesis, the information of multi-dimensional signals is investigated for GMTI. Several methods are provided to settle some key problems of GMTI, which can be summarized as follows:1、In traditional uniform radar array systems, a relatively longer baseline would result in a well minimum detectable velocity(MDV), but the maximum un-ambiguous velocity and the detectable velocity range would be reduced. To resolve this contradiction, specific optimal designs of the multi-frequency radar array system is provided, on one hand, multi-frequencies are utilized for detectable velocity range extension and velocity ambiguity resolution; on the other hand, the increased spatial sampling by non-uniform array is able to improve the detection performance. Incorporated with the non-uniform array, the devised multi-frequency system would obtain a remarkable detection performance. Two specific designs of the multi-frequency radar array system are provided. The cascade one uses the minimum redundancy theory for array optimization and selects the frequencies according the idea of non-overlap detectable range. The joint one searches the frequencies as well as the phase center positions simultaneously, aiming at to obtain a best detectable range. Remarkable improvement against the traditional single frequency uniform array is demonstrated with theoretical analysis as well as numerical simulation.2、In multi-polarization radar array system, the combination of polarization filtering and space-time filtering provides a tremendous potential to improve the MDV for GMTI. However, the improved filtering domain would worse the problem caused by the high dimension of STAP. A comprehensive analysis of polarization-space-time three dimensional processing is provided based on the establishment of signal model. Meanwhile, the major factors that effect the processing, i.e., the polarization coherence of clutter, polarization number and the polarimetric difference between the clutter and target, are pointed out. For casecade processing, a conclusion can be derived that the first filter should be processed on the dimenion which owns a larger target and clutter difference. The conclusion is verified with numerical simulation.3、In nonhomogeneous environment, a small sample set support is offten encoutered and a subsequent detection degradation is unavoidable. In this thesis, we propose a clutter covariance matrix estimation algorithm using multi-polarized data in the polarimetric radar system, which can mitigate this problem with an enlarged training sample set. Based on the space-time signal model, the clutter snapshots for different polarizations share a common spectral structure theoretically is validated. Then, the maximum likelihood estimations of clutter covariance matrixes with multi-polarized training samples are deduced under Gaussian and Non-Gaussian statistic. Finally, the performance improvement of the proposed method with limited training samples is demonstrated with simulated data.4、Secondary data selection algorithm, which eliminates the contanminated samples from the secondary data set, is commonly required for adaptive processing in GMTI. The performances of the traditional sample selection algorithms are limited since they works on the space-time dimension only. Since the polarimetric analysis would provide a well classification results in SAR images, it is feasible to improve the data selection with the assistant of polarimetric data. Improved classification results are derived based on an extended polarization-space two-dimensional Wishart iteration. Then we use the classification results to instruct secondary sample selection with the recursive generalized inner product(GIP) algorithm. The robustness of the proposed method as well as the target detection performance improvement is demonstrated with simulated results.5、Prolate Spheroidal Wave Function(PSWF) method could improve the target detection performance for space-time adaptive processing(STAP) in nonhomogeneous environment. However, its performance may be remarkbaly reduced in situation of system parameter error. In this thesis, we correct the system parameter with clutter spectrum analysis. Since contaminated samples contained in the secondary data set have detrimental impact on this spectrum analysis, the traditional sample selection method of generalized inner production(GIP) is combined with PSWF method, and then a bi-iterative scheme is proposed. Several vital issues such as how to estimate the parameter with real data and why the precision of covariance matrix could be improved during the iteration are analyzed. In the end, the validity of the proposed algorithm is substantiated by practical and simulation results.