海杂波噪声中小目标的特征分析与检测方法研究

发布时间：2018-02-21 18:26

  本文关键词： 海杂波 微弱信号检测 混沌 噪声 分形　出处：《南京信息工程大学》2016年硕士论文　论文类型：学位论文

【摘要】：海杂波受到诸如海浪、海风和潮汐等环境因素影响,具有类似噪声的特性,是典型的非平稳信号。在雷达对海进行观测时,海杂波大量尖峰会严重影响雷达对海监测效果,由于小目标的雷达反射截面积很小,容易淹没在海杂波和噪声中。传统检测方法存在精度低、泛化性差和实时性欠佳的问题。如何从海杂波背景下有准确、可靠地发现小目标,是当前雷达信号处理领域的研究重点。本文研究了混沌混合信号的噪声抑制和利用问题,分别提出基于EMD方差特性的混沌信号自适应去噪算法和基于粒子群优化的自适应随机共振检测方法。研究了海杂波在FRFT域下的分形特性,在单尺度和高尺度条件下,分别提出了基于阶数自适应的小目标检测方法和基于多重分形的小目标检测方法。引入分形聚类方法筛选海杂波数据,用于支持向量机的选择性集成学习,提出了基于自适应分形聚类的微弱信号检测方法。具体的研究重点如下：基于经验模态分解理论,研究了不同混沌系统分解分量的方差特性,发现噪声导致总分解层数的增加和分解分量方差最大值所在层数增大,确定需要去噪处理的分量层数,结合提升小波的优势,提出一种基于EMD方差特性的混沌信号自适应去噪算法。采用Lorenz、Chen系统和实测海杂波雷达数据进行实验。结果表明,在低噪条件下,比传统小波去噪方法均方误差降低30%以上,信噪比提高15dB-3.5dB,可在保留有用信号的基础上有效地去除海杂波中的噪声,提高海杂波数据质量。根据随机共振系统中利用噪声能量增强待测微弱信号的特性,研究二维Duffing振子参数对随机共振输出信噪比的影响。利用粒子群算法全局优化的特点,对二维Duffing振子三个参数进行优化,提出一种基于粒子群的自适应随机共振的微弱信号检测方法,将自适应随机共振微弱信号检测问题转化为参数并行寻优问题,实测海杂波数据实验表明,输出信噪比提升明显,能有效地从海杂波背景中检测到微弱周期信号。研究了海杂波在FRFT域下的分形特点,采用分数布朗运动建模,推导证明其受阶数和尺度的双重影响。根据FRFT补偿雷达信号速度和加速度补偿的特点,在单尺度下提出基于阶数自适应的海杂波小目标检测方法,有效地提高了检测门限,比分形时域检测方法提高26.3%。在高尺度下提出基于高尺度多重分形的小目标检测方法,发现在负高尺度上纯海杂波与目标单元差异明显,两种方法均较好地解决海清变化对小目标检测的干扰。进一步引入分形聚类法筛选海杂波数据,用于提高支持向量机训练效率,提升了海杂波背景下的微弱信号检测性能。本文通过对海杂波噪声背景下小目标的特性进行分析研究,结合经验模态分解、随机共振和分形等理论,提出了海杂波去噪和微弱信号检测方法,较好地缓解了海情对目标信号的干扰,对海面小目标的识别和海面安全监测具有一定的理论意义和实际应用价值。
[Abstract]:Sea clutter is affected by wind and tidal waves, effects of environmental factors such as characteristics, similar to noise, is a typical non-stationary signal. In the observation of radar on sea, sea clutter large spikes will seriously affect the radar on sea monitoring effect, because the RCS Target is very small, easy to drown in the sea clutter and noise. The traditional detection method of low precision, poor generalization capability and poor real-time problems. From the background of sea clutter accurately, reliably detect small targets, is a current research topic in radar signal processing field. This paper studies the use of noise suppression and chaotic signal problems. Put forward adaptive chaotic signal EMD variance characteristic denoising algorithm based on particle swarm optimization and adaptive stochastic resonance detection method based on the research. The fractal features of sea clutter in FRFT domain, the single scale and high scale. Under the proposed respectively small target detection method based on adaptive order and based on small target detection method of multi fractal. The fractal clustering method for the screening of sea clutter data for selective ensemble learning support vector machine is proposed, weak signal detection method based on adaptive fractal clustering. Focus on specific experience as follows: Based on the theory of modal decomposition, variance decomposition characteristics of components of different chaotic systems, found that noise causes the total decomposition layers increase and decomposition of variance maximum value number increases, determine the need for denoising the component layer, combined with the advantages of wavelet transform, proposed an adaptive denoising algorithm based on chaotic signal based on EMD variance characteristics. Using Lorenz, Chen system and sea clutter radar experimental data. The results show that in low noise conditions, compared with the traditional wavelet denoising method of mean square error The difference is reduced by more than 30%, to improve the signal-to-noise ratio of 15dB-3.5dB, can effectively remove the noise in sea clutter based on retaining the useful signal and improve the quality of sea clutter data. According to the noise energy using stochastic resonance system to enhance properties of weak signal to be measured, influence of two-dimensional Duffing oscillator parameters on ratio of stochastic resonance the output signal to noise. Using the characteristic of particle swarm algorithm for global optimization of two-dimensional Duffing oscillator, three parameters are optimized, a weak signal detection method of adaptive particle swarm optimization based on stochastic resonance, the problem of adaptive stochastic resonance weak signal detection into a parameter optimization problem, the measured sea clutter data experiments show that the output signal-to-noise ratio can significantly enhance the weak periodic signal detection effectively from sea clutter. The sea clutter fractal characteristics of wave in FRFT domain, using the fractional Brown motion model, The derivation is greatly influenced by the order number and scale. According to the characteristics of FRFT radar signal compensation speed and acceleration compensation, adaptive order sea clutter based on small target detection method is proposed in the single scale, effectively improve the detection threshold than the fractal time domain detection method of high 26.3%. of small target detection in high scale based on multi fractal in high scale, found in the negative high scale pure sea clutter and the target unit is significantly different, two methods to solve the interference of Hai Qing change on small target detection based on fractal clustering method. Further screening of sea clutter data is used to improve SVM training efficiency, enhance the weak the performance of signal detection in sea clutter background. Based on the characteristics of small target in sea clutter wave noise background analysis, combined with empirical mode decomposition, stochastic resonance and fractal theory, put forward The method of sea clutter denoising and weak signal detection has better relieved the interference of the sea situation to the target signal, and has certain theoretical significance and practical application value for small target recognition and sea surface safety monitoring.

【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN957.51
，

本文编号：1522550