【作者】 常玉林；

【导师】 周智敏；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2009， 博士

【摘要】 不同于高频SAR/GMTI系统,低频UWB SAR/GMTI系统的波束角很大,运动目标信号能量散布于较大角度范围内,输入信杂噪比很低;波长较长,对运动目标径向速度的敏感程度较低,这些因素的存在,给低频UWB SAR/GMTI系统下的运动目标检测带来了很大困难。本文采用结合了“长相干积累间隔”(Long Coherent Processing Intervals,长CPI)思想和“空时自适应处理”(Space Time Adaptive Processing, STAP)思想的长CPI STAP技术,作为低频UWB SAR/GMTI系统的运动目标检测手段。长CPI STAP技术是传统STAP技术的推广,它突破了传统STAP“一个CPI之内运动目标和载机之间的距离走动不超过一个距离采样单元”的限制,可利用更长CPI,达到更高的输出信杂噪比和更强的运动目标检测能力。本论文对长CPI STAP的理论和方法进行了系统的研究。在长CPI STAP的基本理论方面,本文分析了运动目标和地杂波的长CPI STAP空时谱,指出当观测角度变化范围较大时,运动目标的长CPI空时谱则可能为非线性的;提出了长CPI STAP基本模型,和传统短CPI STAP模型仅包含空间-慢时间二维采样不同,本文的长CPI STAP基本模型包含空间维-慢时间维-快时间维的三维采样,可适应运动目标的跨距离单元走动;分析了长CPI STAP降维处理和短CPI STAP降维处理的差异,指出了长CPI STAP降维滤波器的设计原则。本文第三章将长CPI STAP基本模型变换到频率-多普勒域,设计了频率-多普勒域的最优长CPI STAP滤波器,并根据不同频率-多普勒单元的杂波统计独立性,对其进行降维处理。降维处理后,频率-多普勒域长CPI STAP分为“频率-多普勒域局部STAP”和“频率-多普勒域局部STAP结果相干积累”两大步骤。其中,“频率-多普勒局部STAP”对每一个频率-多普勒单元进行,“频率-多普勒域局部STAP后结果相干积累”要求先取出一定速度对应的频率-多普勒STAP结果,然后采用ω-k成像算法完成相干积累,这两个步骤均可设计为与运动目标位置参数无关的形式,从而可对场景中所有位置的运动目标同时进行长CPI STAP。在以上基础上,提出了频率-多普勒域长CPI STAP运动目标检测流程,并对其性能进行了预测。本文第四章将长CPI STAP基本模型变换到图像域,设计了图像域最优长CPI STAP滤波器,该模型包含空间-方位向-距离向三维,能够适应运动目标可能在UWB SAR图像上的散焦。然后利用不同像素间的杂波单元的统计独立性,设计了基于单像素和基于多像素的降维长CPI STAP模型。降维处理后,图像域长CPI STAP可分为“图像域局部STAP”和“图像域局部STAP结果相干积累”两大步骤,且这两个步骤均可设计为与运动目标位置参数无关的形式。其中,图像域局部STAP对每一个像素的多通道观测值进行。本文提出了理想情况下、存在非均匀杂波情况下、通道失配情况下的图像域局部STAP技术。“图像域局部STAP结果相干积累”要求取出一定速度对应的图像域局部STAP结果,然后将其作二维FFT变换到二维波数域进行散焦补偿。在此基础上,提出了基于图像域长CPI STAP的运动目标检测流程,并对其性能进行了预测。本文第五章将长CPI STAP基本模型变换到多子孔径图像域,设计了多子孔径图像域最优长CPI STAP滤波器,并利用不同子孔径杂波的统计独立性,将多子孔径图像域最优长CPI STAP降维表示为“子孔径STAP”和“子孔径STAP结果相干叠加”两个独立的步骤。这两个步骤均可表示为与运动目标位置参数无关的形式。提出了基于多子孔径图像域STAP的多分辨运动目标检测方法,该方法通过综合不同数目的子孔径,进行多分辨运动目标检测。该方法能够结合短CPI和长CPI的优点,可兼顾快速运动目标检测“检测速度快”和慢速运动目标检测“输出信杂噪比高”的要求。在第三至五章,基于同一批半实测三通道低频超宽带SAR数据,对这三种长CPI STAP方法分别进行了验证,并对其性能进行了评估。结果表明:频率-多普勒域长CPI STAP方法,除了可适应运动目标沿慢时间维的跨距离单元走动之外,还可适应运动目标沿通道维的距离走动,并能够利用系统的大相对带宽,消除天线稀疏放置引起的盲速和重复检测现象;图像域长CPI STAP方法对非均匀杂波环境具有更好的适应能力;而基于多子孔径图像域长CPI STAP能够进行多分辨运动目标检测,可以很好的兼顾了短CPI STAP时间分辨率高和长CPI运动目标检测性能好的优点。基于长CPI STAP直接对运动目标参数进行估计时,参数估计精度受到搜索参数的选取间隔或者基线长度的限制。本文第六章对基于长CPI STAP的运动目标参数估计技术进行了研究,提出了一种完整的长CPI STAP运动目标参数估计方案。该方案结合了长CPI STAP、自聚焦、ATI的思想,可在采用较稀疏的搜索间隔和较短基线的情况下,达到较高的参数估计精度。更多还原

【Abstract】 Unlike high frequency SAR/GMTI systems, low frequency UWB SAR/GMTI system has large beamwidth, which makes moving target signal disperse to a large angle range and results in a low input SINR. Besides, its long wavelength is less sensitive to range velocity of the moving target. So it is difficult to detect moving target using UWB SAR/GMTI system.This paper invested the long CPI STAP technique for moving target detection using low frequency UWB SAR/GMTI. Unlike traditional short CPI STAP, the long CPI STAP can endure the range walk between range bins, thus can achieve higher output SINR and better performance of moving target detection.The long CPI space-time spectrum of moving target and clutter is first analyzed, which can serve as the basic principle of long CPI STAP. Then a basic long CPI STAP model is proposed, which involves three-dimension samples from space, slow time, and fast time.By transforming the basic long CPI STAP model to frequency-Doppler domain, the optimal frequency-Doppler domain long CPI STAP filter is introduced In Chapter 3 of the dissertation. Then the dimension of the model is optimal reduced based on the statistic independence characteristic of clutter in different frequency-Doppler bins. After this operation, the frequency-Doppler domain long CPI STAP can be reduced to two steps:“local STAP”and“coherent integration of local STAP”in frequency-Doppler domain. The former step is done in each frequency-Doppler unit, while the latter one requires selecting the output of frequency-Doppler local STAP corresponding to certain velocity and then coherent integration is realized applyingω-k imaging algorithm. Both the two steps can be designed as forms independent of position parameter of the moving target, which make it possible to calculate the output long CPI STAP simultaneously for all moving target in different position of the scene. Moreover, a moving target detection flow for frequency-Doppler domain long CPI STAP is proposed.By transforming the basic long CPI STAP model to image domain, the optimal image domain long CPI STAP filter is introduced in Chapter 4 of the dissertation. This model involves three-dimension of space-azimuth-range, thus can adapt the defocusing phenomenon of the moving target in UWB SAR image. Based on the statistic independent characteristic of clutter in different pixels, the optimal image domain long CPI STAP filter can be reduced to single-pixel or multi-pixel reduced-dimension filters.After this operation, the image domain long CPI STAP contains two steps as“local STAP”and“coherent integration the output local STAP”in image domain. Both the two steps can be designed as forms independent of position parameter of the moving target. The local STAP is done in each pixel along its multi-channel observations. We propose image domain local STAP technology in ideal environments, nonhomogeneous environments.“Coherent integration local STAP in image domain”requires is done by first selecting the output of image domain local STAP corresponding to certain velocity, and then transforming to two-dimension wavenumber domain to compensated defocusing. Building on the ground work above, a moving target detection flow for image domain long CPI STAP is proposed. Its predictive performance is also given.By transforming the basic long CPI STAP model to subaperture-image domain, the optimal subaperture- image domain long CPI STAP filter is proposed in Chapter 5 of the dissertation. Using the statistic independent characteristic of clutter in different subaperture, this model has its reduced-dimension form, which contains two independent steps as“subaperture STAP”and“coherent integration the output of subaperture STAP”. Both the two steps can be designed as forms independent of position parameter of the moving target. A multi-resolution moving target detection method based on subaperture image domain STAP is proposed, which combines different number of subapertures to realize moving target detection. The method combines the merits of short CPI for fast moving target detection and long CPI for slow moving target detection.Throughout Chapter 3 to Chapter 5, the three long CPI STAP methods are validated and their performances are valuated based on the same half-real triple-channel low frequency UWB SAR data. The results show that the frequency-Doppler domain long CPI STAP method can be applied to the conditions that the moving target’s range walk across range cell both in slow time domain and channel-dimension. The method also exploits the large relative bandwidth of the system to eliminate blind velocity and repeated detection caused by sparse placed antennae. The image domain long CPI STAP method has better adaptability for nonhomogeneous clutter environment. The subaperture image domain long CPI STAP method gets the ability of multi-resolution moving target detection, having the advantages of high time-resolution from short CPI STAP and good moving target detection performance from long CPI STAP.When estimate the moving target’s parameters directly based on long CPI STAP, the estimation accuracy is limited by sample interval of search parameters and length of baseline. In Chapter 6 of the dissertation, we research on the parameter estimation technology and propose a complete method for moving target parameter estimation based on long CPI STAP. Combining long CPI STAP with the concept of auto-focusing and ATI, this method can reach high parameter estimation accuracy when using sparse research interval and short baseline.更多还原