连续波雷达地面慢速目标检测与分类

发布时间：2018-06-13 21:40

本文选题：连续波雷达 + 目标检测　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：连续波体制雷达由于很好地解决了脉冲体制的距离盲区问题,并具有良好的反隐身、抗背景杂波和抗干扰能力,已广泛应用于地面监视雷达。而在地面目标监视中,行人和车辆等慢速目标通常是监视的重要对象,它们的运动和趋势关系着态势的走向,相关检测技术已经得到研究人员的广泛关注。同时,随着现代科技快速发展和各国军事战略的技术要求,雷达的功能不仅仅局限于探测和测距方面。目标的分类能给态势分析带来更多的信息,已经成为雷达,特别是地面监视雷达另一个重要的实际应用方向。因此,连续波雷达慢速运动目标检测与分类方法是目前雷达信号处理的研究热点。本文结合实际应用需求,对连续波体制雷达的目标检测与分类进行了系统的研究,所取得的研究成果为:(1)对线性调频连续波雷达的工作原理和信号预处理方法进行了研究。推导了多周期的信号模型以及差拍信号的频域响应特性,为目标的检测分析奠定了基础。根据目标的特性和需求的区别,详细论述了动目标显示、CLEAN和杂波图等杂波抑制方法,实现了杂波和目标的区分,为后续的慢速目标检测奠定了基础。(2)深入研究了目标检测过程的距离速度耦合问题及其解决方法。针对锯齿波调频,利用线性调频信号的参数估计方法和模糊速度估计的方法,实现了目标距离和速度的准确反演;针对三角波调频,利用目标频谱对称性,在传统的频谱配对方法的基础上,介绍了基于动目标检测(Moving Target Detction,简称MTD)的频域配对和提出了时频配对方法,从而实现了运动目标的准确配对,进而精确地获得了目标的真实距离和速度。(3)深入研究了人与车辆等慢速目标的微多普勒谱特征。建立了行人和车辆等慢速目标的数学模型,并利用实测数据验证了建模的有效性。然后,结合仿真和实测数据,分析了行人和车辆等慢速目标微多普勒特性的差异。(4)针对目标的微多普勒特征的差异性,提出了一种人与车辆连续波雷达分类新方法。该方法以目标时频谱图的灰度共生矩阵的能量、熵值、对比度和自相关的均值和方差等统计量为特征,利用主分量分析进行特征降维。在得到的本征特征量中选取部分样本作为分类器的训练样本,输入支持向量机分类器进行分类。仿真和实测数据处理结果表明该方法可以稳健地实现人、车谱图数据的监督分类。
[Abstract]:Continuous wave radar (CWR) has been widely used in ground surveillance radar because of its good anti-stealth, anti-background clutter and anti-jamming ability, because it can solve the problem of blind area of pulse system. In ground target surveillance, slow targets such as pedestrians and vehicles are usually the important objects of surveillance, their movement and trend are related to the trend of the situation, the related detection technology has been widely concerned by researchers. At the same time, with the rapid development of modern science and technology and the technical requirements of the military strategy of various countries, the functions of radar are not limited to detection and ranging. Target classification can bring more information to situation analysis and has become another important practical application direction of radar, especially ground surveillance radar. Therefore, the detection and classification of slow-moving targets in continuous wave radar is a hot topic in radar signal processing. In this paper, the target detection and classification of continuous-wave radar is systematically studied according to the practical application requirements. The research result is: 1) the principle and signal preprocessing method of LFM CWR are studied. The multi-period signal model and the response characteristics of beat signal in frequency domain are derived, which lays a foundation for target detection and analysis. According to the difference of target characteristics and requirements, the clutter suppression methods such as moving target display clear and clutter graph are discussed in detail, and the distinction between clutter and target is realized. It lays a foundation for the following slow target detection. (2) the distance and velocity coupling problem in the process of target detection and its solution are studied in depth. For zigzag wave frequency modulation, the parameter estimation method of linear frequency modulation signal and the method of fuzzy velocity estimation are used to realize the accurate inversion of target distance and velocity, and for triangular wave frequency modulation, the symmetry of target frequency spectrum is used. Based on the traditional spectrum pairing method, this paper introduces the frequency domain pairing based on moving target detection (MTD) and proposes a time-frequency pairing method to realize the accurate pairing of moving targets. Finally, the real distance and velocity of the target are obtained. (3) the characteristics of micro-Doppler spectrum of the slow target, such as human and vehicle, are studied in depth. The mathematical model of slow target such as pedestrian and vehicle is established, and the validity of the model is verified by the measured data. Then, based on the simulated and measured data, the difference of micro-Doppler characteristics between pedestrian and vehicle is analyzed. A new method of continuous wave radar classification is proposed for the difference of micro-Doppler characteristics. This method is characterized by the energy, entropy, contrast and autocorrelation mean and variance statistics of the gray level co-occurrence matrix of the target spectrum. The principal component analysis is used to reduce the dimension of the feature. Some samples are selected as the training samples of the classifier and the support vector machine classifier is input to classify the eigenvalues. The simulation and measured data processing results show that the proposed method can be used to monitor and classify the data of human and vehicle spectra.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.51

【共引文献】