【作者】 游伟；

【导师】 何子述；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2013， 博士

【摘要】 天波超视距雷达利用电离层对电磁波的反射，实现对远距离海面舰船、空中目标的检测、参数估计及跟踪。天波超视距雷达探测距离可以不受地球曲率的限制，实现对800-3500公里远海面目标探测。正是由于天波雷达的这一优势，使得其在远程警戒、早期预警、机场监视、缉毒走私、海洋遥感等领域都得到了广泛的研究与应用。尽管天波雷达有许多优势，但是由于天波雷达信号传输环境复杂，其在目标检测尤其是慢速舰船目标检测方面仍然存在许多挑战。本文围绕这些问题展开深入研究，其主要工作和贡献包括：1、由于瞬态干扰持续时间较短，幅度强，会掩盖目标频谱使得目标无法检测。本文在充分研究瞬态干扰时域与频域特性的基础上，提出一种基于S变换的瞬态干扰抑制算法。该算法将时域回波变换到时间-多普勒域，从而可以在时间-多普勒二维平面上对瞬态干扰进行检测和定位，并在时域对其有效抑制。2、基于时域挖除与线性预测恢复的瞬态干扰抑制技术，其性能通常受到线性预测性能的影响。本文提出一种基于自适应高斯线性调频变换(AGC, AdaptiveGaussian Chirplet Transform)的算法。该算法将原始回波分解为一系列AGC基函数的和，再利用瞬态干扰和回波中其余分量在时域与频域的不同特点直接将瞬态干扰分量消去。该算法不需要利用线性预测技术，从而避免了线性预测技术对抑制效果的影响。同时，文中引入遗传算法来实现AGC基函数分解的多维搜索。3、对于慢速舰船目标检测而言，由于慢速舰船目标径向速度较低，其多普勒频率(Doppler frequency)通常靠近一阶海杂波。为了检测慢速舰船目标，需要有较长的相干积累时间。电离层的上下运动会给长时间相干积累的回波信号带来相位噪声。该噪声使得海杂波的频谱扩展从而进一步掩盖舰船目标，使舰船检测更加困难。电离层的这一影响通常称为电离层污染。考虑到海杂波频谱的扩展根本原因在于杂波的瞬时频率随每个脉冲发生变化，本文提出一种基于复数能量检测(Complex Energy Detection, CED)的算法来对杂波瞬时频率进行估计并将其用于电离层污染的校正。仿真结果表明该算法能够在较大污染的情况下取得较好的污染校正效果。对于天波雷达的电离层污染校正，针对三次相位信号模型，提出一种新的三次相位信号瞬时频率估计算法。该算法将海杂波分成多个较短的数据段，然后在每个数据段求取脉冲点的瞬时频率，最后将所得到的瞬时频率平滑得到每个脉冲的瞬时频率。将得到的频率用于对污染的校正。仿真结果表明，该算法频率估计精度比文献中提出的多项式相位变换(Discrete Polynomial phase Transform, DPT)算法要高，并且可以在较大污染的情况下实现校正。4、对于慢速舰船而言，其频谱通常位于两个一阶谱峰之间，则这类目标的检测更加困难。因为除了一阶杂波外，二阶杂波也可能会影响目标的检测。本文在分析海杂波特点的基础上，首先对杂波协方差矩阵进行子空间分解，在此基础上，将MVDR(Minimum Variance Distortionless Response)理论用于实现舰船目标的有效检测。该算法可以抑制二阶连续杂波，适用于检测位于一阶峰之间但不被一阶峰掩盖的目标。5、对于飞机、弹道导弹等高速目标，目标可能出现机动的情况，即这类空中目标会在一段时间内做加速运动。对这类目标而言，传统的距离-多普勒处理会造成信噪比的损失，从而降低目标的检测概率。这类目标的回波可以建模为多项式相位信号。本文基于此模型，提出一种新的多项式相位信号参数估计算法。与已提出的DPT算法相比，该算法能够在更低的信噪比条件下实现机动目标的参数估计和运动补偿。在给定的仿真参数下，当信噪比为3dB时，算法对机动目标的频谱增益可达5dB以上。6、对空中目标而言，其飞行高度往往也是感兴趣的参数之一。比如，高度信息可以用于对目标进行分类。本文基于天波雷达空中目标的微多径模型，提出一种基于子空间的高度估计算法。该算法利用观测数据的噪声子空间与仿真数据的信号子空间进行匹配来获得目标的高度估计。给出了仿真结果，仿真结果表明在给定仿真参数下所提算法比最大似然算法有更好的稳健性。更多还原

【Abstract】 The skywave over-the-horizon (OTH) radar detects ships and aerial targets over thesea surface, estimates the parameters and tracks these targets by utilizing the reflectedelectromagnetic wave from the ionosphere. The detection range of the skywave OTHradar is not limited by the curvature of the earth, but can detect targets over the seasurface up to800-3500km. Due to this advantage, it has been widely studied andapplied in remote surveillance, early warning, airport surveillance, anti-smuggling andcounter drug, ocean remote sensing, etc.Although the skywave OTH radar has many advantages, it still has manychallenges in target detection especially slow ship target detection due to its complexsignal propagation environment. This dissertation mainly concerns these aspects. Themain research results and contributions include:1. The transient noise has very short duration time but with very strong magnitude.The target spectrum will be masked by the transient noise which makes the targetcannot be detected. On the basis of fully research on the time domain characteristic andfrequency domain characteristic of the transient noise, a transient noise excisionalgorithm based on the S transform is proposed. The algorithm transforms the timedomain return signal onto the time-Doppler domain, which makes the transient noisecan be detected and localized on the two dimensional plane accurately, thus the transientnoise can be excised in time domain effectively.2. The performance of the time domain excision and linear pridiction algorithms isgenerally affected by the linear prediction techniques. In this dissertation, an algorithmbased on the adaptive Gaussian chirplet (AGC) transform is proposed. The algorithmdecomposes the original returned signal into the sum of a series of AGC basis functions,and then the transient noise can be subtracted directly by utilizing the differentcharacteristics between the transient noise and the other signal components in both timeand frequency domains. No linear prediction technique is needed in this algorithm, thusthe adverse effect to the excision can be avoided. Meanwhile, genetic algorithm (GA) isintroduced for the multidimensional search of the AGC basis decomposition.3. For the slow ship target detection, its Doppler frequency is generally close to thefirst-order sea clutter due to its small radial velocity. In order to detect the slow ship target, a relatively long integration time is necessary. The movement of the ionospherewill result in phase noise to the long time integrated returned signal. The noise broadensthe spectrum of the sea clutter and masks the ship target, thus makes the detection moredifficult. The ionospheric effect is generally referred to as the ionospheric contamination.The root cause of the spectrum spread of the sea clutter is the instantaneous frequencyof the clutter changes with each pulse. Taking this into account, an algorithm based onthe complex energy detection (CED) is proposed to estimate the instantaneousfrequency of the clutter and applied to correct the ionospheric contamination.Simulation results show that the proposed algorithm can work and achieve goodperformance in large ionospheric perturbations.For the skywave radar decontamination, based on the cubic phase modeling, a newcubic phase signal instantaneous frequency (IF) estimation algorithm is proposed. Thealgorithm divides the sea clutter into several short data segments, and then estimates theIF of each pulse for the data segments. The final IF of each pulse can be calculated byaveraging the IF of each overlapped data segments. The derived IF is then utilized forthe decontamination. Simulation results show that the IF estimation accuracy is betterthan the Discrete Polynomial phase Transform (DPT) algorithm, and can workeffectively under large ionospheric perturbations.4. For slow ship target, its spectrum generally lies between the two first-orderclutter peaks, and then the target detection is more difficult. Besides the first-orderclutter, the second-order clutter can also influence the detection of the target. Based onthe analysis on the characteristic of the ocean clutter, the covariance matrix of the clutteris firstly decomposed into subspaces. Then the MVDR theory is used to detect the shiptarget. The algorithm can suppress the second-order continuous clutter, thus it can beused to detect the target which lies in between the first-order clutter but not masked.5. For high-speed target, such as aircraft and ballistic missile, target maneuvering isprobably to happen. That is, the aerial target will accelerate during the time duration.For this kind of target, the traditional Range-Doppler processing will lead tosignal-to-noise ratio (SNR) lossing, thus reduce the probability of target detection. Thetarget return signal can be modeled as a polynomial phase signal. Based on this model, anew polynomial phase signal parameter estimation algorithm is proposed. Comparingwith the DPT algorithm, the proposed algorithm can achieve maneuvering targetparameter estimation and movement compensation under lower SNR. Under the specified simulation parameters, when the SNR is3dB, the proposed algorithm obtains5dB or more in the spectrum gain.6. For the aerial target, its flying altitude is also an interested parameter for theskywave radar. For example, the altitude can be used for classifying the aerial targets.Based on the micro-multipath model of the aerial target for the skywave radar, asubspace altitude estimation algorithm is proposed in this dissertation. The algorithmderives the altitude of the target via matching the noise subspace of the observed datawith the signal subspace of the simulated data. The simulation results are presented, andthey show that the proposed algorithm is more robust than the maximum likelihood(ML) algorithm under the specified simulation parameters.更多还原