【作者】 王成；

【导师】 郁文贤；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2006， 博士

【摘要】 雷达信号层融合成像是一种新兴的雷达成像处理技术,它是多传感器信号层融合与雷达成像技术相结合的产物。它旨在通过对来自不同时间、空间或频率上的雷达观测信息进行相干融合,以提高目标散射中心参数估计精度并获得更高的雷达成像分辨率,从而有利于目标识别与图像理解等后期处理。传统的单雷达成像系统受单传感器信号带宽与观测相干积累时间的约束,雷达成像分辨率有限。随着雷达技术和信息处理技术的不断发展,利用多雷达观测信息进行信号层融合成像成为克服单雷达成像系统局限的一种有效方式。基于多雷达观测信息的有效融合,可以明显改善雷达成像分辨率,提高目标散射中心参数的估计精度。本文针对逆合成孔径雷达成像的应用背景,研究了多雷达信号层融合成像处理中的一些基础问题以及不同融合方式下的多雷达信号层融合成像算法,包括多频带雷达信号层融合成像的理论框架、信号幅相补偿、多频带多雷达信号层融合的高分辨距离成像算法、稀疏多子带雷达信号层融合的超分辨距离成像算法、二维信号融合超分辨ISAR成像算法等。首先,提出了多频带多雷达信号层融合成像的理论框架,在研究多频带多雷达信号融合回波模型的基础上,建立了实际信号处理中统一的幅相误差模型;详细分析了残余相位误差与雷达间隔距离、目标尺寸、观测几何等因素的相互关系以及残余相位误差对一维距离成像的影响,给出了三种降低残余相位误差的雷达布站方式;研究了多频带雷达信号层融合成像的基本原理,分析了实际信号处理中非均匀采样对幅相补偿和信号融合一维距离成像的影响,并提出了基于插值的均匀重采样处理方法。其次,根据信号幅相误差补偿模型,利用信号模型的子空间特性,对于可重叠的频带观测数据分别提出了基于快速Root-MUSIC和基于ESPRIT-LS的幅相误差补偿参数估计算法;为了减少频域数据长距离外推预测引入的预测误差对幅相补偿的影响,提出了基于熵最小准则的幅相补偿方法,该方法无需重叠频带数据,有效地减少了频域数据外推预测的距离。再次,从已有电磁模型的分析出发,根据雷达回波信号的非平稳特性,提出了基于非平稳时间序列处理的雷达信号融合高分辨距离成像算法,比较了基于Prony模型、ARIMA模型、TVAR模型的多频带雷达信号融合高分辨一维距离成像方法,结果表明基于TVAR模型的融合成像算法具有更好的融合成像效果;为了更好地表示回波信号中的不同成分,在分析距离像多分辨率性质的基础上,将多分辨分析与非平稳建模相结合,提出了DT CWT+TVAR的信号融合高分辨距离成像算法,在不同的尺度空间上对信号进行非平稳建模,较好地刻画了信号中的局部特征然后,针对稀疏多子带观测下雷达信号带宽间隔过大,已有信号融合高分辨距离成像方法不再适用的情况,建立了幅相补偿和散射中心参数估计的统一模型,研究了如何通过空间谱估计技术来进行参数估计,详细分析了空间平滑处理对参数估计的影响,并提出了两种特殊的空间平滑处理方式来简化参数估计;从简化计算的角度考虑,通过构造特殊的空间导向矢量矩阵,将线性相位误差和固定相位误差的估计解耦,利用信号的旋转不变特性,进一步提出了一维搜索的ESPRIT超分辨成像算法;针对超分辨成像算法,分别从Rayleigh准则和Cramér-Rao限的角度出发,定性、定量分析了雷达信号融合对成像分辨率的改善作用,推导了单雷达超分辨成像和雷达信号层融合超分辨成像中散射中心参数估计的Cramér-Rao限,并分析比较了各种因素对距离成像分辨率的影响。最后,分析了单雷达成像与双雷达信号融合ISAR成像在波数空间上的差异,针对共面成像情况,研究了双雷达信号层融合ISAR成像的预处理技术,提出了消除线性相位误差、简化初始时刻雷达视线夹角估计的参考距离匹配技术,并提出了双雷达融合下的目标转动参数估计方法;基于双雷达信号层融合成像模型,提出了一种半解耦的双雷达观测融合超分辨ISAR成像算法;为了更充分地利用多频带信息,更好地改善纵向分辨率,进一步提出了双雷达观测融合的联合超分辨ISAR成像算法。仿真结果表明所提出的两种算法对目标散射中心参数的估计精度均高于单雷达成像估计结果,并具有较好的稳健性。更多还原

【Abstract】 Radar signal level fusion imaging (RSLFI) is an emerging radar imaging technology. It is a combination of multi-sensor signal level fusion and radar imaging technology. It aims at improving the estimation precision of scattering center parameters and the resolution of radar imaging through the coherent fusion of the radar observations information from the different time, space and frequency. It is also in favor of target recognition and image comprehension. It is known that the resolution of traditional monostatic radar imaging system is restricted by the signal band width and observations coherent accumulate time. With the development of radar and information processing technology, the signal level fusion imaging based on the multi-radar observations becomes an effective method of overcoming the limitations of the monostatic radar imaging system. The information fusion based on the multi-radar observations can obviously improve the radar imaging resolution and the scattering center parameters estimation precision.Aiming at the background of inverse synthetic aperture radar (ISAR) imaging, this dissertation studied some basic problems and imaging algorithms in multi-radar signal level fusion imaging processing, including the theoretic frame, amplitude-phase compensation, high resolution range profile (HRRP) formation based on the multiple frequency band radar observations, supper-resolution range profile formation based on the sparse multiple subband radar observations and two-dimensional signal fusion supper-resolution ISAR imaging.The theoretic frame of multiple frequency band and multi-radar signal fusion imaging was presented firstly in the dissertation. Based on the research on the returned signal model of multi-radar fusion, a uniform amplitude-phase error (APE) model was founded. The relationship between residual phase error and the baseline of the two radars, the target size and the observation geometry were analyzed. The effects of residual phase error on the range profile formation were discussed and three modes of radar distribution were presented in order to reduce the residual phase error. The essential theory of multiple frequency band and multi-radar signal fusion imaging was also studied. After analyzing of the effects of the non-uniform sampling on amplitude-phase compensation and range profile formation based on the signal fusion, the uniform resample processing method was proposed.Based on the amplitude-phase error compensation (APEC) model of returned signal, two algorithms were presented. The parameters estimation of the algorithms utilizes the subspace characteristic of signal model and requires the overlap frequency observations of the two radars. Because the longer the signal extrapolated length is, the bigger the signal errors aroused will be. In order to reduce the effect of the signal extrapolated error on the APEC, an APEC algorithm based on the entropy-minimization principle was proposed. The algorithm needn’t the overlap frequency observations and can effectively decrease the signal extrapolated length.With the analysis for the known scattering models and the non-stationary characteristic of radar signal, a HRRP formation method of multiple frequency band radar signal fusion was proposed based on the non-stationary time sequence processing. Compared with the HRRP formation algorithm based on the Prony model and autoregressive integrated moving average (ARIMA) model, the time-variant autoregressive (TVAR) model based algorithm has better fusion imaging result than others. In order to express the different components of the radar returned signal, the multi-resolution characteristic of range profile was analyzed and a HRRP formation algorithm based on the dual tree complex wavelet transform (DT CWT) and TVAR was presented. The algorithm builds the different non-stationary time sequence model for different scale signal based on the multi-resolution analysis and non-stationary modeling. It can better express the local characteristic of the returned signal.Under the sparse multiple subband observations condition, the known HRRP formation algorithms based on the signal fusion can not be applied because the blank between the two radar frequency bands is great. So a uniform parameter model for APEC and scattering center parameters estimation was presented. The estimation method of the model parameters based on the space spectrum estimation was studied. The effect of the spatial smoothing on the parameters estimation was analyzed and two spatial smoothing methods were proposed to reduce the complexity of parameters estimation. For the convenience of parameters estimation, a supper-resolution imaging algorithm based on the one dimension search and the estimation of signal parameters via rotational invariance technique was presented. The estimation of linear phase error and constant phase error were decoupled through constructing the special spatial steer vector matrix in the algorithm. The improvement function of radar signal level fusion on the resolution of radar imaging was analyzed with Rayleigh criterion and Cramer-Rao bound (CRB) for supper-resolution imaging algorithm. The CRB of scattering center parameters estimation were derived in monostatic radar imaging and multi-radar signal level fusion imaging. The effects of various factors on the resolution were analyzed and compared.Finally, the diversity of monostatic radar imaging and dual radar signal level fusion imaging was analyzed in wavenumber space. With the coplanarity imaging condition, the preprocessing technology of dual radar signal fusion imaging was studied. A reference distance matching method was proposed to remove the linear phase error and simplify the estimation of the angle between the two radars’ line-of-sight. A new target rotation parameter estimation method was also put forward for the dual radar signal fusion. Based on the signal level fusion imaging model of the dual radar, a half-decoupled supper-resolution ISAR imaging algorithm with dual radar observations fusion was presented. Subsequently, a unite supper-resolution ISAR imaging algorithm with dual radar observations fusion was proposed in order to take full advantage of multiple frequency band to improve range resolution. The simulation results show that the two algorithms have much better estimation precision and robustness than monostatic radar imaging algorithm for the target scattering center parameters.更多还原