基于时频分析的雷达目标识别技术研究

【作者】 俞悦；

【导师】 盛卫星；

【作者基本信息】 南京理工大学 ， 通信与信息系统， 2012， 硕士

【摘要】 本论文主要针对不同微动运动的目标,开展了基于时频分析的雷达目标识别技术的研究。本文的主要工作包括：1.研究目标的微动特性的时频图。选用三类目标圆锥、柱锥、球锥作为研究模型,通过GRECO建模软件计算出带有径向运动翻滚、旋转、进动等微动的RCS,分别研究在这几种目标运动形式下目标的时频分析图。在MATLAB软件平台上,分别对短时傅里叶变换(STFT)、Wigner-Ville变换、伪Wigner-Ville变换和小波变换等这几种时频分析方法进行了比较,验证了使用伪Wigner-ville变换具备更高的频率分辨性能,且具有更好的频移适应性。2.提出了直接使用波形熵作为特征分类的方法对目标进行识别,而对于不同径向运动速度的目标提出了通过平移波形熵图的方法来提取稳定的特征,并分析了在不同时频分析方法不同运动状态下的目标的波形熵提取结果,验证了波形熵作为特征分类的可行性。3.完成了对目标识别算法的性能分析。从信噪比、时频分析方法、运动形式、分类器等四个方面验证了该识别算法的性能。更多还原

【Abstract】 In this thesis, focused on the targets in air with micro-movement, radar target recognition based on time-frequency analysis is studied. The main research work includes:1. Using time-frequency method to analysis object with Micro-Motion is studied. In the modelling software of GRECO, three targets with simple structure, cone, combination of sphere and cone, combination of cylinder and cone are taken as research object to caculate their dynamic RCS echo signals in different micromotion ways, such as roll, rotation, precession with the radial motion. Three types of time-frequency analysis methods, including short-time Fourier transform (STFT), Wigner-Ville transform, pseudo-Wigner-Ville transform and wavelet transform, are compared on the platform of MATLAB software. It is verified that pseudo-Wigner-ville transform can have a higher frequency resolution performance and a better adjustability with frequency shift.2. A classification method by using the waveform entropy feature is proposed. For different radial velocity of the target, the method of moving the waveform entropy diagram in the horizontal direction to extract the stable features is introduced. With the waveform entropy extraction analysis result of the target in different motion state by using different time-frequency analysis methods, waveform entropy as the feasibility of the feature classification is confirmed.3. The analysis of target classification algorithm performance is completed. From the four aspects of signal-to-noise ratio, frequency analysis methods, form of micromotion, classifiers, the performance of classification algorithm is confirmed respectively.更多还原